PROJECT REPORT ON

BUDGET MANAGEMENT SYSTEM

Submitted to

Department of Computer Applications in partial

fulfillment for the award of the degree of

BACHELOR OF COMPUTER APPLICTIONS

Batch (2022-2024)

Submitted by

Prashant Mamgain

Enrollment Number: GE-22212967

Under the Guidance of Mr. Harendra Singh Negi

GRAPHIC ERA DEEMED TO BE UNIVERSITY DEHRADUN

MAY -2024
 CANDIDATE’S DECLARATION

I hereby certify that the work presented in this project report entitled
“_______________________________________________________” in partial fulfilment of the
requirements for the award of the degree of Bachelor of Computer Applications is a Bonafede work
carried out by me during the period of APRIL 2024 to MAY 2024 under the supervision of
_______________, Department of Computer Application, Graphic Era Deemed to be University,
Dehradun, India.

This work has not been submitted elsewhere for the award of a degree/diploma/certificate.

Name and Signature of Candidate

This is to certify that the above-mentioned statement in the candidate’s declaration is correct to the
best of my knowledge.

Date: ____________ Name and Signature of Guide

Signature of Supervisor Signature of External Examiner

 HOD

Acknowledgement

First of all, my sincere and wholehearted gratitude to my advisor, [Mr. Harendra Singh Negi], for
their invaluable guidance, support, and encouragement throughout the duration of this project.
Their expertise and insights have been instrumental in shaping the direction and outcome of my
work.

I would also like to extend my thanks to the Mentor, whose resources and assistance were crucial
in completing this project. Their dedication to providing a supportive learning environment has
been greatly appreciated.

I am profoundly grateful to my family and friends for their constant support, patience, and
understanding. Their encouragement has been a source of motivation and strength throughout this
journey.

Furthermore, I would like to acknowledge the authors and researchers whose works have
significantly contributed to the foundation of my project on handwritten digit recognition using
neural networks. Their pioneering research and publications provided essential knowledge and
inspiration.

Lastly, I would like to thank [Mr. Harendra Singh Negi], which was critical to the completion of this
project. Without their support, this project would not have been possible.

Thank you all for your contributions and support.

 Candidate Name and Signature
CHAPTER 1

INTRODUCTION

1.1 ABOUT THE ORGANIZATION

GRAPHIC ERA DEMMED TO BE UNIVERSITY

Graphic Era (Deemed to be University), formerly known as Graphic Era Institute of Technology, is a
prestigious educational institution located in Dehradun, Uttarakhand, India. Let me share some interesting
details about this university:

1. History and Accreditation:

o Graphic Era Institute of Technology was established in 1997 with a focus on providing
world-class education in cutting-edge technologies, professional development, critical
thinking, and quality research.
o In 2008, it was accorded the status of Deemed University under Section 3 of the UGC Act,
1956.
o The university has been accredited by the National Assessment and Accreditation Council
(NAAC) with a grade of “A+”. o It is also approved by the All India Council for Technical
Education (AICTE) and the University Grants Commission (UGC).
o Graphic Era University has consistently featured in the Top 100 Universities of India
according to the National Institutional Ranking Framework (NIRF) established by the
Government of India.
2. Campus and Location:
o The university is situated in the picturesque city of Dehradun, surrounded by the Rajaji
National Park on one end and Clement Town Cantonment on the other.
o The serene environment provides an ideal setting for holistic student development.
3. Academic Offerings:
o Graphic Era offers a wide range of undergraduate (UG), postgraduate (PG), and doctoral
programs across various domains, including:
 Engineering
 Computer Application
 Management
 Commerce
 Life Sciences
  Food Technology
 Hospitality
 Paramedical
 Humanities and Social Sciences
4. Distinguished Faculty: o The university boasts a highly qualified teaching faculty, including
professors educated from premier institutions such as IITs, NITs, IISc Bangalore, and IIMs.
 SYNOPSIS

The project work is a practical experience of the knowledge one has. The documentation
leads a way to the concept to present the thinking and the upgradation of various techniques into the
project. This project entitled “BUDGET MANAGEMENT SYSTEM” is a practical project based
on some trends of computer science. Every day the world is searching new techniques in the field of
computer science to upgrade the human limitations into machines to get more and more accurate
and meaningful data. The way of machine learning and artificial intelligence has no negative slop it
has only the slop having positive direction. This project is a very basic idea of those concepts. This
project deals with the very popular learning process called Neural Network. There are various ways
by which one can achieve the goal to a desired output, but in machine learning Neural network gives
a way that machine learns the way to reach the output.

This project has come through the concepts of statistical modeling, the computer vision
and machine learning libraries which includes a lot of study about these concepts. I tried to lead
these project to the end of some updated techniques, upgradation and application of some new
algorithms . This project has a good explanation and this project can be enhanced further into
some complex applications of machine learning.

(iv)
 TABLE OF CONTENT

1. INTRODUCTION………………………………………………………………………………...
1

1.1 About the organization………………………………………………………………. 1

1.2 About the Project…………………………………………………………………….. 2

2. PROBLEM DEFINITION AND FEASIBILITY ANALYSIS ………………………………... 3

2.1 Problem Definition…………………………………………………………………... 3

2.2 Need of the software………………………………………………………………… 4

2.3 Feasibility Analysis ………………………………………………………………..... 5

3. SOFTWARE REQUIREMENT SPECIFICATION…………………………………………… 6

3.1 System Requirement ………………………………………………………………………………. 6

3.2 Tools and Technologies…………………………………………………………………………... 7

4. MODEL DETAILS………………………………………………………………………………..
8

4.1 Types of Models…………………………………………………………………………………... 8

4.1.1 Features of models ……………………..……………………………………………..

8

4.1.2 Security Model…………………….……………………………………..

17

4.2 The Neural Network Model…………………………………………………………………...... 18

4.2.1 Neural network

………………………………………………………………………… 18
 4.2.2 Training Part……………………………………………………………………….. 20

4.2.3 Testing Part…………………………………………………………………………..22

(v)

5. FLOW CHARTS …………………………………………………………………………………………….

24
6. TESTING………………………………………………………………………………………………………….. 29

7. CONCLUSION AND FORSEEABLE ENHANCEMENT…………………………………………...

31

7.1.Conclusion…………………………………………………………………................ 31

7.2.Forseeable Enhancement……………………………………………………………...32

8. REFERENCES…………………………………………………………………………………………………... 33

APPENDIX A-SCREENSHOTS INPUTS …………………………………………………….. 34

APPENDIX B – OUTPUTS………………………………………………………………... 37

(vi)
 LIST OF FIGURES

Figure Figures Page Number

No.

5.1 Image Processing for Training Data 24

Image 25
5.2 Processing for Testing data

5.3 Neural Network model for Training Data 26

Backpropagation 27
5.4 Algorithm(upgraded with Adam Optimiser)

Neural Network Model 28

5.5 for Testing Data
 (vii)
1.2 ABOUT THE PROJECT

A budget management system is a tool or framework designed to help individuals, organizations, or

governments plan, monitor, and control their financial resources efficiently. It ensures that funds are
allocated appropriately, expenses are tracked, and financial goals are met. Here's an overview of the
key aspects of a budget management system:

Purpose of a Budget Management System

1. Financial Planning: Helps allocate resources to achieve specific goals.

2. Expense Tracking: Monitors spending to ensure adherence to the budget.
3. Decision Support: Provides insights for making informed financial decisions.
4. Goal Setting: Aligns financial planning with short- and long-term objectives.
5. Accountability: Encourages transparency and accountability in financial management.

Core Features

1. Budget Creation: Allows users to define budgets for specific categories (e.g., rent, salaries,
marketing).
2. Expense Tracking: Logs actual expenditures and compares them to budgeted amounts.
3. Reporting Tools: Provides visualizations like graphs, pie charts, and reports to analyze
financial performance.
4. Alerts and Notifications: Warns users about overspending or remaining balances.
5. Integration: Connects with accounting systems, payroll, or bank accounts for seamless data
syncing.

Benefits

1. Prevents Overspending: Helps control expenses by adhering to set limits.

2. Enhances Savings: Ensures surplus funds can be channeled into savings or investments.
3. Improves Efficiency: Reduces manual errors and saves time through automation.
4. Increases Transparency: Makes financial data accessible and understandable.
5. Supports Strategic Planning: Facilitates long-term financial stability.

Applications

1. Personal Budget Management: Apps like Mint, YNAB (You Need A Budget), or PocketGuard
assist individuals with their finances.
2. Business Budgeting: Tools like QuickBooks or FreshBooks help businesses manage cash flow
and project budgets.
3. Government Finance Management: Large-scale systems designed for national or local
government budgets.

Challenges

1. Data Accuracy: Requires accurate input to ensure reliable outcomes.

2. Scalability: May need updates or adjustments for growing needs.
3. User Adoption: Employees or individuals might resist switching to a new system.
4. Cost: Advanced systems might involve significant setup and maintenance costs.
 CHAPTER 2

PROBLEM DEFINITION AND FEASIBILTY ANALYSIS

2.1 PROBLEM DEFINITION

The problem definition for a budget management system focuses on identifying the key financial
challenges faced by individuals or organizations and the gaps in current methods of managing
budgets.

Identified Problems:

1. Lack of Organization:
o Many individuals or organizations struggle to track their income and expenses
systematically.
o Manual processes (e.g., spreadsheets or paper records) are prone to errors and
timeconsuming.
2. Overspending and Budget Deviations:
o Difficulty in adhering to set budgets often leads to overspending or misallocation of
funds.
o Real-time tracking of spending is not possible with conventional methods.
3. Limited Insights and Decision Support:
o Without detailed reports and insights, financial decisions are often based on incomplete
data.
o Inadequate forecasting leads to poor long-term planning.
4. Inefficiency and Lack of Automation:
o Manual entry of expenses and income can lead to inefficiencies.
o Repetitive tasks (e.g., categorizing expenses) consume significant time.
5. Scalability Issues:
o Current tools or processes might not scale with increasing complexity, such as
organizational growth or additional financial variables.
6. Lack of Transparency and Accountability:
o In organizations, the absence of transparent budget tracking mechanisms can lead to
financial mismanagement or fraud.

2.2 NEED OF THE SOFTWARE

A budget management system is essential for ensuring effective financial planning and resource
allocation. It addresses several challenges faced by individuals, organizations, and governments in
managing their finances. Here's why a budget management system is needed:
 1. Efficient Financial Planning

• Helps allocate financial resources effectively to various needs and priorities.

• Ensures alignment of expenditures with short-term and long-term financial goals.
• Enables forecasting of future financial needs and revenues.

2. Expense Control and Oversight

• Tracks and monitors spending in real time to prevent overspending.

• Provides visibility into financial leaks or unnecessary expenses. Alerts users
when budgets for specific categories are exceeded.

3. Transparency and Accountability

• Ensures accurate financial reporting and compliance with regulations.

• Facilitates transparency in financial transactions, reducing the risk of fraud or mismanagement.
• Promotes accountability among stakeholders, such as employees, managers, or family members.

4. Decision Support and Insights

• Provides detailed reports and analytics to support informed decision-making.

• Offers insights into spending patterns, income trends, and cash flow.
• Helps identify cost-saving opportunities and areas for financial optimization.

5. Automation and Efficiency

• Reduces manual effort in tracking, categorizing, and analyzing finances.

• Automates recurring tasks, such as bill payments or generating financial reports.
• Saves time and minimizes errors compared to manual methods.

6. Scalability for Growing Needs

• Adapts to the increasing complexity of financial management as needs evolve (e.g., business
expansion, more income streams).
• Supports additional users, categories, and integrations with other systems as required.

2.3.1 Technical Feasibility

A feasibility analysis evaluates whether developing and implementing a budget management system is
practical and beneficial.

1. Technical Feasibility:
 • Availability of Technology:
o Modern software development tools and frameworks (e.g., cloud-based platforms, APIs,
mobile integration) make it feasible to design user-friendly systems.
o Integration with financial services like banks or payment gateways is achievable.
• Skills and Expertise:
o Availability of skilled developers, UI/UX designers, and financial analysts can ensure
the system's success.

2. Economic Feasibility:

• Cost of Development:
o Development costs include software design, database setup, cloud hosting, and
maintenance.
o Cost-effectiveness depends on the scope (e.g., individual, enterprise, or
governmentlevel system).
• ROI (Return on Investment):
o Reduction in financial errors and better decision-making will lead to cost savings. o
Scalability ensures long-term benefits, making initial investment worthwhile.
3. Operational Feasibility:

• Ease of Use:
o User-friendly interfaces and automation can encourage adoption by individuals and
organizations.
• Employee or User Training:
o Training programs can address resistance to new systems.
• Adoption Rate:
o Systems offering significant advantages over current methods will likely see high
adoption rates.

4. Legal and Compliance Feasibility:

• Data Security and Privacy:

o Adherence to data protection regulations (e.g., GDPR, HIPAA) is critical, especially for
financial systems.
• Audit and Compliance:
o Incorporating features for financial audits and regulatory compliance is necessary for
enterprise users.

5. Time Feasibility:

• Development Timeline:
o A robust system can be developed in phases, prioritizing key features for a Minimum
Viable Product (MVP).
• Deployment and Updates:
o Cloud-based systems can enable quicker deployment and seamless updates.
 CHAPTER 3 SOFTWARE REQUIREMENT SPECIFICATION (SRS)

An SRS document defines the functional, non-functional, and technical requirements for developing a
Budget Management System. Below is a comprehensive outline:

1. Introduction
1.1 Purpose

The purpose of this software is to provide individuals, businesses, and organizations with a tool to
plan, monitor, and control their financial resources efficiently. The system will streamline budgeting
processes, track expenses, generate reports, and enable informed decision-making.

1.2 Scope The system will support financial planning by allowing users to create, manage, and
monitor budgets.
• Features include real-time tracking of expenses, alerts for budget overruns, and insightful
financial reporting.
• Applicable for individual, small business, or organizational use. Available on web and
mobile platforms for accessibility.
1.3 Intended Audience and Use

• Users: Individuals, families, small businesses, and organizations.

• Stakeholders: Developers, financial planners, management teams, and investors.

1.4 Definitions, Acronyms, and Abbreviations

• SRS: Software Requirements Specification.

• MVP: Minimum Viable Product.
• API: Application Programming Interface.

2. Functional Requirements
2.1 User Authentication

• Secure login and registration using email, username, or social media accounts.
• Role-based access for users (admin, manager, individual user).

2.2 Budget Creation and Management

• Allow users to create monthly, quarterly, or yearly budgets.

• Add categories (e.g., rent, salaries, utilities) and allocate amounts.
• Edit or delete existing budgets.

2.3 Expense Tracking

 • Input expenses manually or import from linked bank accounts.
• Categorize expenses for better organization.
• Provide an overview of budget vs. actual spending.

2.4 Reporting and Analytics

• Generate summary and detailed reports (e.g., pie charts, bar graphs).
• Provide spending trends and future projections.
• Export reports in formats like PDF and Excel.

2.5 Alerts and Notifications

• Notify users of upcoming bills, due dates, or budget overages.

• Real-time alerts for significant spending deviations.

2.6 Multi-Platform Support

• Responsive web application for desktops. Mobile applications for iOS and Android.

2.7 Data Security

• Encrypt sensitive user information (e.g., account details).

• Role-based access controls to prevent unauthorized actions.

2.8 Integration with Financial Systems

• Support integration with banking systems via APIs for real-time updates. Link with
accounting tools like QuickBooks or Xero.
2.9 Backup and Restore

• Automatic data backup to cloud storage.

• Restore previous budgets or data snapshots.

3. Non-Functional Requirements
3.1 Performance

• The system should handle at least 1,000 concurrent users without performance degradation.
• Transactions should be logged and updated within 2 seconds.

3.2 Scalability

• The system should support an increasing number of users and transactions as the organization
grows.

3.3 Usability

• Intuitive interface for both tech-savvy and non-technical users.

• Provide in-app tutorials and tooltips for ease of use.
3.4 Availability

• Ensure 99.9% uptime with failover systems to handle server crashes.

• Scheduled maintenance should be notified to users in advance.

3.5 Security

• Implement multi-factor authentication for user login.

• Adhere to data protection laws (e.g., GDPR, CCPA).
• Ensure encryption for all data in transit and at rest.

3.6 Compatibility

• Compatible with major browsers (Chrome, Firefox, Edge, Safari).

• Mobile apps should work on devices running iOS 12+ and Android 8.0+.

4. System Requirements
4.1 Hardware Requirements

• Client Side:
o Desktop: Minimum 2 GHz processor, 4 GB RAM, and 500 MB free storage.
o Mobile: Minimum 2 GB RAM, 50 MB free storage.
• Server Side:
o Minimum 4-core processor, 16 GB RAM, 1 TB storage. o
Cloud-based hosting with scalable resources.

4.2 Software Requirements

• Frontend: o HTML5, CSS3, JavaScript (React, Angular, or Vue).

• Backend: o Node.js, Python (Django/Flask), or Java (Spring Boot). Database:
 PostgreSQL, MySQL, or MongoDB.
• Operating System:
o Server: Linux-based (Ubuntu or CentOS). o
Client: Windows, macOS, iOS, and Android.
• APIs:
o Open Banking API, payment gateway APIs (Stripe, PayPal).

5. Assumptions and Constraints

• Users will provide accurate financial data for tracking.

• Initial development will focus on an MVP for individual users, with business and enterprise
features rolled out later.
• Internet connectivity is required for cloud-based operations.

6. Future Enhancements

• Machine learning for predictive analytics and personalized recommendations.

• Support for multiple currencies and languages.
• Voice-command integration for hands-free operations.
 o
CHAPTER 4 MODEL DETAILS

4The Budget Management System can be modeled using a layered architecture, modular components,
and data-driven processes. Below is a detailed breakdown of the model design:

1. Architectural Model
1.1 Layered Architecture

1. Presentation Layer (Frontend):

o Purpose: Interface for users to interact with the system. o Components:
 Dashboards for financial summaries.
 Forms for budget creation, expense input, and report generation.
 Visualization tools like charts and graphs.
o Technologies:
 Web: React, Angular, or Vue.js.
 Mobile: Flutter or React Native.

2. Application Layer (Backend Logic):

o Purpose: Processes requests from the frontend and manages business logic.
o Components:
 Budget creation and modification logic.
 Expense categorization and tracking engine.  Alert and notification
management.
o Technologies:
 Node.js, Django, or Spring Boot for API development.

3. Database Layer:
o Purpose: Stores, organizes, and retrieves financial data. o Components:
 User profiles and credentials.
 Budget and expense records.
 Audit trails and historical data.
o Technologies:
 Relational DBs: PostgreSQL or MySQL.  Non-relational DBs: MongoDB
(for flexible data structures).
4. Integration Layer:
o Purpose: Facilitates communication with external systems. o Components:
 APIs for bank account integration (e.g., Open Banking).
 Payment gateway APIs (e.g., Stripe, PayPal).
 Accounting software APIs (e.g., QuickBooks).
 2. Component Model

2.1 User Management Module

• Features:
o User registration, login, and role-based access. o
Multi-factor authentication for enhanced security.
• Data Model:
o User ID, Name, Email, Role, and Password (encrypted).

2.2 Budget Management Module

• Features: o Define budgets by category (e.g., food, rent, utilities).

o Set periodic budgets (daily, monthly, yearly).
• Data Model:
o Budget ID, User ID, Category, Allocated Amount, and Period.

2.3 Expense Tracking Module

• Features:
o Record and categorize expenses manually or via bank sync.
o Auto-detection of recurring expenses.
• Data Model:
o Expense ID, Date, Category, Amount, Payment Method.

2.4 Reporting and Analytics Module

• Features: o Generate visual and tabular reports (e.g., pie charts, trend lines).
Export reports in formats like PDF or Excel.
• Data Model: o Aggregated budget and expense data by category, period, and user.

2.5 Notification Module

• Features:
o Real-time alerts for overspending or nearing budget limits.
o Reminders for bill payments and upcoming expenses.
• Data Model: o Notification ID, User ID, Type, and Status (read/unread).

3. Data Flow Model

1. User Interaction:
 o
o Users input budgets, add expenses, or request reports via the frontend.
o Requests are validated and sent to the backend for processing.
2. Backend Processing:
o Business logic is applied (e.g., checking if expenses exceed budgets).
o Data is stored or retrieved from the database.
3. Notifications and Reports:
o Backend triggers alerts based on user-defined thresholds. o Processed data is sent back
to the frontend for visualization or export.

4. UML Diagrams
4.1 Use Case Diagram

• Actors: User, Admin.

• Use Cases:
o Create Budget. o
Add Expense. o
Generate Report.
o Set
Notifications.

4.2 Class Diagram

• Classes and Relationships:

o User → Budget (1-to-Many). o
Budget → Expense (1-to-Many). o
User → Notification (1-to-Many).

4.3 Sequence Diagram

Example: Adding an

Expense

1. User enters expense details.

2. System validates and saves the data.
3. Updated summary is displayed.

5. Process Model
5.1 Budget Workflow

1. User logs in.

 2. Creates a new budget by specifying categories and amounts.
3. System saves the budget and displays it on the dashboard.

5.2 Expense Workflow

1. User inputs an expense or imports it from linked accounts.

2. System categorizes and checks against the budget.
3. If the expense exceeds the budget, a notification is triggered.

5.3 Reporting Workflow

1. User selects a time range or specific category.

2. System aggregates data and generates reports.
3. Report is displayed or exported.

5. Technical Model 6.
6.1 Cloud Deployment Model

• Features:
o Scalable architecture using AWS, Google Cloud, or Azure.
o Load balancing to handle high traffic.
• Components: o Virtual machines for hosting the application. o Cloud storage for secure data
backup.

6.2 Security Model

• Features:
o Data encryption (AES-256). o Role-
based access control. o API rate
limiting to prevent misuse.

Purpose of Neural Network in Budget Management System

1. Expense Prediction:
o Predict future expenses based on historical data and user spending patterns.
o Identify trends to help users allocate budgets effectively.
2. Anomaly Detection:
o Detect irregular spending patterns or potential fraudulent transactions. o Alert users
about unusual activities that deviate from normal behavior.
3. Categorization of Expenses:
Automatically classify expenses into predefined categories using transaction
descriptions. o Improve accuracy and save users time.
4. Financial Advice:
 o
o Provide recommendations for better savings or investment opportunities. o
Suggest optimal budget allocations based on user behavior and goals.
5. User Personalization:
o Learn user-specific financial habits to customize budget goals, alerts, and advice.

Neural Network Architecture

1. Input Layer

• Accepts input features such as:

o Historical expenses (e.g., last 6 months' data). o Income details. o
Expense categories. o Transaction metadata (e.g., time, location,
amount). o User preferences and goals.

2. Hidden Layers

• Multiple dense layers for feature extraction and learning complex patterns.
• Techniques:
o Recurrent Neural Networks (RNNs) or Long Short-Term Memory
(LSTM): Handle sequential data like monthly expenses over time. o
Convolutional Neural Networks (CNNs): Process metadata and
transaction descriptions for categorization.

3. Output Layer

• Provides predictions or classifications such as:

o Next month's expenses (regression output).
o Anomaly score for each transaction (classification output). o
Recommended budget allocation or advice (categorical outputs).

Example Neural Network Model

Model 1: Expense Prediction

1. Input Features:

o Historical spending (time-series data). o Seasonal factors (e.g.,

holiday periods).
o User income and savings rate.
2. Network Type:
o LSTM layers to process sequential data.
o Dense layers to integrate contextual features.
3. Output:
o Predicted total expenditure for the next period.
Model 2: Anomaly Detection

1. Input Features:

o Transaction details (amount, time, category).

o User spending patterns.
2. Network Type:
o Autoencoder to learn normal spending patterns and flag deviations.
3. Output:
o Anomaly score for each transaction.

Model 3: Expense Categorization

1. Input Features:

o Transaction descriptions and amounts.

2. Network Type:
o CNN to analyze text descriptions and classify transactions.
3. Output:
o Predicted category for each transaction.

Neural Network Architecture

1. Input Layer

• Accepts input features such as:

o Historical expenses (e.g., last 6 months' data). o Income details. o
Expense categories. o Transaction metadata (e.g., time, location,
amount). o User preferences and goals.

2. Hidden Layers

• Multiple dense layers for feature extraction and learning complex patterns.
• Techniques:
o Recurrent Neural Networks (RNNs) or Long Short-Term Memory
(LSTM): Handle sequential data like monthly expenses over time. o
Convolutional Neural Networks (CNNs): Process metadata and
transaction descriptions for categorization.

3. Output Layer

• Provides predictions or classifications such as:

o Next month's expenses (regression output).
o Anomaly score for each transaction (classification output). o
Recommended budget allocation or advice (categorical outputs).
 o

Example Neural Network Model

Model 1: Expense Prediction

1. Input Features:

o Historical spending (time-series data). o

Seasonal factors (e.g., holiday periods).
User income and savings rate.
2. Network Type:
o LSTM layers to process sequential data.
o Dense layers to integrate contextual features.
3. Output:
o Predicted total expenditure for the next period.

Model 2: Anomaly Detection

1. Input Features:

o Transaction details (amount, time, category).

o User spending patterns.
2. Network Type:
o Autoencoder to learn normal spending patterns and flag deviations.
3. Output:
o Anomaly score for each transaction.

Model 3: Expense Categorization

1. Input Features:

o Transaction descriptions and amounts.

2. Network Type:
o CNN to analyze text descriptions and classify transactions.
3. Output:
o Predicted category for each transaction.

3.1. Hierarchy

At first,to know the contour hierarchy,how to find contours of a binary image?

region, hi=cv2.findContours(edged. copy (), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

There are three arguments in cv2.findContours(); first one is the source image;
second is contour retrieval mode; third is the contour approximation method.
 Contour hierarchy deals with contour retrieval mode and contour
approximation method. There are various types of retrieval and approximation
method.

findcontours() retreives all the contours in an image,but sometimes or most of the

times for handwritten character set some contours stays inside one contour or connected with
another contour.The outer contour called Parent,inner connected contour called Child and
the connected contours called Sibling.This type of relationship is Hierarchy of contours.

Problem of Contour Hierarchy

Now the main theory about contour is the contour is totally dependent on variance on the
pixel,like whenever it gets a variance it applies the contour on it. So, in case of handwritten digit
extraction it’s not efficient because inside an image it will consider more and more sibling contours
and then total family while only the supreme of that family is needed.This thing totally can’t be
considered because inside one contour one child is there but comparing to another it can be parent
also.Inside one contour there comes another contour also that will extract more than one contour
from a character.in fig 6,for “Q” it will cut 4 contour 1 outside and three inside contour.For “8” it
will cut the same also.This is problematic.

3.2. Upgradation in Bounding Box

In this paper, the problem in the concept of bounding box will be upgraded.Mainly,
Bounding box is one type of box which can be rectangle or oval shaped or circle by which an
object in a particular image or video can be recognised.Mainly, for handwritten character
recognition rectangle boxes can be used to describe the target location box.Bounding box gives
the output in a 4 valued list.
So what has to be done is to delete those bounding boxes which are inside and to keep only those
bounding boxes which are outside and giving the boundary to the characters. For Fig 12, 13 only
the outside box has to be kept and not the inside. All the bounding boxes values can be obtained
and put into one matrix or 2-D list.

3.2.2 Duplicate or Bounded bounding box deletion

There can be many bounding boxes which are fully bounded by another bigger
bounding box that need to be deleted because there can be pixel variation inside a character
but that will not be considered.Only outside bounding box will be considered.

The picture has the upper left point(x,y),width and height of every bounding box has
been stored in our matrix.Take every two lists from the matrix and check if those bounding box
has any bounded bounding box or not.

Check if the one bonding box (x,y) point inside another .If inside then check the width+x
and height+y if inside bounding box is fully inside.If yes then delete it or add another one into
mainmatrix.After all take only distinct values. Then sort the matrix as those contours need to be
obtained and to fit it into the distinct positions.Like the contour which has been created for 8(fig
12) is residing in a matrix so we have to fit that contour onto 8 only and same for another
characters also.If one bounding box is started inside one but its height or width goes beyond the
bigger one which is mostly outside then inside contour will be there.Only fully bounded inside
contour will be deleted or not been considered.

3.3 Putting contours

After detection of the right contours for the image, in the ideal contour bounding box
recognition the number of bounding box found will be equals to the number of characters in the
image. After getting the bounding boxes it is needed to be sorted with removing the duplicates
and then those bounding boxes on the image setting contours have to be put.

x,y,w,h =matrix[i]
imag = cv2.rectangle(image,(x,y),(x+w,y+h),(0,0,255),2)

The very first attribute on the rectangle function is the image on which one will draw the
boxes. Second attribute (x,y) is the upper left corners of the contour in that 2-D image.Third
attribute will denote the right side bound and the lower bound of the image.Fourth attribute will
be the color of the contour box (r,g,b) and the fifth one will be the thickness of the bounding box.

2. Results

The following Table 4.4 depicts the result of the contour removal normal algorithm and updated
contour removal bounding box based algorithm. The accuracy on the new algorithm has been
calculated and also given there.

Table 4.4 Total upgradation in results after applying bounding box removal

Total Tested Samples Contour Removal Accuracy

Updated Contour
Algorithm Removal Bounding
Box

134 56 119 88.8

Noise Cacellation

120 110 116 96.6

Now after getting all of these results the training data gets trained with the total MNIST dataset

and the real dataset into the neural network model.

4.2. THE NEURAL NETWORK MODEL

The neural network lies with the concept of machine learning.At first it creates a model
like brain unit and then like child it trains that model like brain with many and many datasets,for
my project it is the digits.There are two parts of a neural network model.

1) Training Part
2) Testing Part

Before going into the deep concept of these two parts in the project some ideas on the building
blocks and structural parts of neural network (C.C.Jay.Kuo – 2016, Adit Deshpande –
2016) have been shared. Each and every neural network can be structured in three different parts-
1) Input layer 2)
Hidden layer
 3) Output layer

A neural network is put together by hooking together many of our simple “neurons,” so
that the output of a neuron can be the input of another. For example, here is a small neural
network:

Fig-4.1 Neural Network Model

In this figure 4.1, circles are used to denote the inputs to the network. The circles labeled “+1”
are called bias units, and correspond to the intercept term. The leftmost layer of the network is called
the input layer, and the rightmost layer the output layer (which, in this example, has only one node).
The middle layer of nodes is called the hidden layer, because its values are not observed in the training
set. Say, that our example neural network has 3 input units (not counting the bias unit),3 hidden units,
and 1 output unit.

The types of layers in a neural network can be summarized as follows: 1)

Input Layer
Input variables, sometimes called the visible layer.This layer can be of features or the direct input
from the dataset.

2) Hidden Layer
Layers of nodes between input and output layers . There may be one or more of these
layers.More one create the hidden layer more operation will be there,more weights will be there and
there will be less change due to the value of difference in this layer.This layer is the most important
part of the neural network.Weights change in these layers varies in the returning input.

3) Output Layer
A layer of nodes that produce the output variables and then based on the targets it can
perform iteration.Finally, there are terms used to describe the shape and capability of a neural
network; for example:

Size: The number of nodes in the model.

Width: The number of nodes in a specific layer.

Depth: The number of layers in a neural network.

Capacity: The type or structure of functions that can be learned by a network

configuration. Sometimes called “representational capacity“. Architecture: The specific
arrangement of the layers and nodes in the network.

4.2.1 TRAINING PART

Training the Neural Network

1. Dataset:
o Collect anonymized financial transaction data, including user income, expenses, and
transaction history.
o Data augmentation to handle imbalances in expense categories.
2. Preprocessing:
o Normalize numerical data (e.g., transaction amounts).
o Tokenize and vectorize text descriptions for categorization tasks.
3. Loss Function:
o Regression tasks: Mean Squared Error (MSE) for prediction models. o
Classification tasks: Categorical Cross-Entropy for categorization.
4. Optimization:
o Use Adam or RMSProp optimizers for faster convergence.

4.2.1.1 CREATING THE NEURAL NETWORK MODEL

Now comes the main part of the project ,so how to create a neural network model that
can take an image and run some probabilistic calculation on it which can implement the model
with better accuracy.
1) Input Layer
 The total model trains on the dataset 28*28 pixels .The neural network model depends upon
the number of inputs and the intensity in input layer and the filtered target output calculation in
output layer. One dimensional array of the intense dataset is needed into this model and then
sequentially give the total MNIST dataset and out retrieved dataset into the neural. One needs to
flatten and sequential the total input into the neural.

2) Hidden Layer

The hidden layer part is the most efficient part in the neural network.It is the part where the
total calculation and the backpropagation like learning algorithm takes place.Inthis project 128
hidden layers have been applied by taking the data from the input layer. In inner input layer and
dense hidden layer MAXRELU function has been applied as an activation function and in the output
layer the softmax activation function has been used (Softmax output function, Jefferey Hinton).

What Activation Function actually does?

Activation function controls the boundary of the data like normalisation .The
normalisation function are rejected in the hidden layer because in the hidden layer the data or
the values gets dynamically changed.There are various types of Activation functions.

a) Sigmoid- Y=1/(1+exp(-x))
b) Tangent- y=(exp(x)-exp(-x))/(exp(x)+exp(-x))
c) MaxReLu- y=max(0,x)
d) Softmax- y=exp(x)/(summation(exp(x)) etc.

In this project the max relu and softmax because this gives really good result tested

and better than others like fig 4.2.Total experiment is based on result no extra theory applied

but the convergence of the sigmoid and the tangential curve to the verified output scenario.
 Fig – 4.2 Types of activation functions over graph

As the linear,logistic(sigmoid) and tangential is too much deflected for stagnant inputs
and touches the curve in 1 and 0 immediately and never comes back practically ,these 3 are
rejected.Sigmoid gives a very good result but practically max relu and softmax is better for
Handwritten digit recognition.

Bias input are rejected to contain the actual shaped data from the main flatten and
sequential data. Learning rate has been given very low 0.002 to make the model more
effective.

A) OPTIMIZER
Optimizers are algorithms or methods used to change the attributes of the neural network such
as weights and learning rate in order to reduce the losses.

There are many types of optimizers.Each and every optimiser has its own method to deal
with the weights and bias inputs.In this project the backpropagation algorithm is modified with
the adam optimizer instead of the gradient descent optimiser. At first gradient descent was set to
be final but then the adam or stochastic gradient descent optimiser is better than anyone as the
inputs from the various pixels are varying in points.

4.2.2. TESTING PART

The testing part comes with some latest real time update in the machine learning part .The
image processing work like webcam access and the other techniques has been discussed before that
is why it has not been discussed here. So this part comes with some ransom upgrade of the technique
which is very small to say but not that much easy to design.

Let’s dive into the model after the total training and taking the figure from the webcam into
that predictive modeled neural network. Set the input pixel values into the numpy array value where
the argmax function will check the range and give the output figure .The range will be done by the
feed forward step of the neural network.No other backpropagation algorithm will be applied because
this is the test to check have the weights updated them to detect that input figure correctly? NO
neural network can come with 100 % accuracy but if it can’t detect most of the test figures then it
must need more training. Result of the model shall be discussed in no time because the model has
been trained more than 23 times to make it better and of course not with the same dataset every time.
Datasets include ones from web and pictures and also handwritten and trained the neural model to
give the correct result in the testing part.

To make things shorter what has been done in this model-

1) Loaded self created data and MNIST dataset into the model

2) Set the numpy array system to take input the kernel along with the data.

3) Create the neural network model setting the input layer and the number of hidden layers and
the output layers along with the activation functions used in different layers. 4) Set the
probabilistic statistical value into the biased dataset into unbiased.

5) Check the target and the output predicted value every time while training the dataset
and set the number of epochs corresponding to the error.

6) Model check with checksum value removal and biased value removal after setting the
weight value by the neural network on its set and set the dataset into the unbiased.

7) Access the webcam and capture the image using python javascript and google colab
display.
8) Check the parameter of probability of the figure after normalisation and noise removal
and scaling. 9) Got highest predicted value?match the figure with the detected from
the neural.

Experimental Testing Result

The Table 4.4 depicts the accuracy count for each iteration in the neural network model.Every time
training takes place it comes with some upgradation on some models.Here 0-9 has been tested 5 times.
Actually 140 datasets have been tested every time. Visualization done by charts in fig 12 Appendix B.
Table 4.4 Experimental Testing Results with the Number of Times Iteration

Digits Test 1 Test 2 Test 3 Test 4 Test 5

(Detected) (Detected) (Detected) (Detected) (Detected)

60 78 110 116 132

0
 50 43 87 97 124
1

72 88 86 114 129
2

23 67 89 109 110
3

45 46 66 78 125
4

56 78 68 90 134
5

76 87 57 104 128
6

45 40 65 99 138
7

8 45 89 95 134 134

55 78 127 130 128

9
 CHAPTER 5 FLOW CHARTS AND FIGURES

6.1 FLOW CHARTS

A flowchart is a type of diagram that represents a workflow or process.A flowchart can also
be defined as diagrammatic representation of an algorithm, a step by step approach.
 Fig 5.2 Image Processing for Testing data

This Flowchart figure 5.2 shows the Image Processing part For the testing data.Testing data
includes the part of the Accessing the webcam and start the video stream on which an image
dataset will be tested.

The neural network model includes processing.Various algorithms are depicted here (fig 5.3)
Fig 5.3 Neural Network Model for Training data
The Backpropagation Algorithm (Fig 5.4) applies Stochastic Gradient
based algorithm. .
Calculate the difference between
Result value

Calculate the
Change backward
In every layer

Initialise learning rate

For each epoch
It will be changing for
every epoch

Calculate the alpha

Inputdelta,hiddendelta
And output delta

3d matrix
hiddenweight
Calculation and
weights
will be changed

Apply
ADAM optimiser to
Add the
Error distribution

Adam optimiser
upgradation applies
Relu and softmax
Predictive analysis
And return
 Fig 5.4 Backpropagation Algorithm (upgraded with ADAM optimiser)

Fig 5.5 Neural Network Model for Testing Data

Neural Network Model gets the prediction from the testing image.Thanks to the Image
processing part for the testing image that actually plots the data in the image into the pixels
calculated the probabilistic values by sending it into the neural network.If matched then it is a
successful one if not then it is an error. May be training will be done with it later. The figure 5.5
depicts the flow of the processes.
 CHAPTER 6 SYSTEM TESTING
System testing evaluates the entire Budget Management System as a complete product to ensure
that it meets functional, non-functional, and performance requirements. The focus is on validating the
system's behavior under various conditions, including normal, edge cases, and failure scenarios.

1. Objectives of System Testing

• Verify that all components of the system (frontend, backend, database, integrations) work
together as intended.
• Ensure the system meets business requirements and user expectations.
• Identify and resolve defects before deployment.

2. Types of System Testing

2.1 Functional Testing

• Verifies that all system features and functionalities are working as expected.
• Key Tests:
o Budget creation and modification. o
Expense tracking and categorization.
o Report generation and export. o
Alerts and notifications functionality.

2.2 Integration Testing

• Ensures seamless communication between system modules.

• Key Tests:
o Frontend-Backend API integration.
o Integration with external systems (e.g., banking APIs, payment gateways). o
Database read/write operations.

2.3 Performance Testing

• Evaluates the system’s response time, throughput, and scalability under various loads.
• Key Tests:
o Load Testing: Handling multiple users concurrently (e.g., 1,000 users).
o Stress Testing: Behavior under extreme conditions (e.g., high transaction volume). o
Response Time: Average time for budget creation or report generation.

2.4 Usability Testing

• Assesses the user experience, including ease of navigation and accessibility.

• Key Tests:
o User onboarding process.
o Responsiveness on different devices (web, mobile). o Intuitiveness of
dashboards and visualizations.
2.5 Security Testing

• Identifies vulnerabilities and ensures data protection.

• Key Tests:
o Authentication (e.g., multi-factor login). o
Authorization (role-based access control). o Data
encryption for sensitive information (e.g., AES-256). o
Protection against SQL injection, XSS, CSRF attacks.

2.6 Compatibility Testing

• Ensures the system works on different platforms, devices, and browsers.

• Key Tests:
o Browser compatibility (Chrome, Firefox, Safari, Edge).
o Mobile compatibility (iOS, Android). o Operating system compatibility
(Windows, macOS, Linux).
2.7 Regression Testing

• Ensures that new updates or bug fixes do not break existing functionalities.
• Key Tests:
o Re-execution of previously passed test cases. o
Testing interdependencies between modules.

2.8 Data Validation Testing

• Confirms the accuracy and integrity of data stored and retrieved.

• Key Tests:
o Correct categorization of expenses. o
Accuracy of budget vs. expense calculations. o
Integrity of exported reports (e.g., no data
loss).

3. Test Scenarios and Test Cases

Scenario 1: User Login

• Test Case 1: Login with valid credentials.

• Test Case 2: Login with invalid credentials and verify error messages.
• Test Case 3: Verify multi-factor authentication functionality.

Scenario 2: Budget Management

• Test Case 1: Create a new budget with valid data.

• Test Case 2: Attempt to create a budget without required fields.
• Test Case 3: Modify and delete an existing budget.

Scenario 3: Expense Tracking

 • Test Case 1: Manually add an expense and verify category assignment.
• Test Case 2: Import transactions from a linked bank account.
• Test Case 3: Add an expense exceeding the allocated budget and verify alerts.

Scenario 4: Report Generation

• Test Case 1: Generate a report for a specified period.

• Test Case 2: Export the report in PDF and Excel formats.
• Test Case 3: Verify accuracy of data in reports.

Scenario 5: Notifications

• Test Case 1: Set and trigger a spending limit notification.

• Test Case 2: Receive reminders for upcoming bills or due dates.

Scenario 6: System Performance

• Test Case 1: Simulate 500 users logging in simultaneously.

• Test Case 2: Stress-test with 10,000 transactions within a short time.

4. Tools for System Testing

4.1 Functional Testing Tools

• Selenium (for automated UI testing).

• Postman (for API testing).

4.2 Performance Testing Tools

• JMeter (load and stress testing).

• Locust (scalability testing).

4.3 Security Testing Tools

• OWASP ZAP (vulnerability scanning).

• Burp Suite (penetration testing).

4.4 Usability Testing Tools

• Crazy Egg (heatmaps for usability).

• UserTesting (collect user feedback).

5. Test Environment Setup

1. Hardware Requirements: o Client-side: Devices with varying specifications for

compatibility testing.
o Server-side: Cloud-based environment with scalable resources.
2. Software Requirements:
o Testing frameworks (e.g., Cypress, JUnit).
o Browsers and emulators for cross-platform testing.
3. Test Data:
o Dummy user profiles with varied budgets, expenses, and income. o Test
scenarios for edge cases (e.g., extreme budget values).

5. Exit Criteria for System Testing

• All critical and high-severity bugs are resolved.

• Functionalities meet the acceptance criteria.
• Performance metrics are within acceptable limits.
• System passes security and usability requirements.
• Regression test suite is executed without failures.
 CHAPTER 7

CONCLUSION AND FORESEEABLE ENHANCEMENT

7.1 CONCLUSION

The Budget Tracking System is a comprehensive solution for individuals and organizations to plan,
monitor, and optimize their finances effectively. By leveraging advanced features such as automated
expense tracking, real-time notifications, and detailed financial reporting, the system enhances user
decision-making and promotes financial discipline.

Through careful design, robust testing, and secure deployment, the system ensures reliability,
usability, and scalability. It caters to various financial management needs, from personal budgeting to
small business expense monitoring, establishing itself as a valuable tool in financial management.

Foreseeable Enhancements

1. Integration of AI and Machine Learning

• Predictive Analytics: Provide future expense forecasts based on historical trends.

• Personalized Recommendations: Suggest tailored financial strategies for savings,

investments, and budget adjustments.

• Expense Categorization Automation: Use natural language processing (NLP) to classify

expenses with higher accuracy.

2. Multi-Currency and Localization Support

• Enable users to manage budgets in multiple currencies with automatic exchange rate updates.
• Add language customization for global accessibility.

3. Voice and Chatbot Support

• Integrate voice commands for hands-free budget management.

• Include AI-driven chatbots for instant financial assistance and query resolution.
4. Gamification Features

• Introduce rewards and challenges to encourage users to stay within their budgets.
• Offer badges or incentives for achieving financial goals, such as monthly savings targets.

5. Financial Goal Setting Allow users to set and track long-term financial goals (e.g., saving for
a vacation or retirement).
• Provide progress indicators and recommendations to help achieve these goals.

6. Advanced Reporting and Insights

• Include benchmarking tools to compare user spending with similar demographics.

• Provide deeper insights into spending patterns and their implications on financial health.

7. Enhanced Integration with External Platforms

• Broader support for linking with third-party financial apps, banks, and payment systems.
• Real-time synchronization with popular accounting software like QuickBooks or Tally.

8. Offline Functionality

• Develop offline capabilities for creating budgets or tracking expenses, with automatic sync once
the internet is available.

9. Sustainability Tracker

• Allow users to categorize and track environmentally friendly expenses.

• Provide insights into spending habits aligned with sustainable practices.

10. Enterprise Features

• Enable advanced budgeting and reporting for larger organizations.

• Add multi-user collaboration tools for teams to manage shared budgets effectively
 CHAPTER 8

REFERENCES

Academic References

1. Books:
o Rosen, H. S., Public Finance, McGraw-Hill, 2021. (Covers budget management
principles and financial planning methodologies.)
o Drury, C., Management and Cost Accounting, Cengage Learning, 2018. (Provides
insights into budgeting in organizational contexts.)
2. Research Papers:
o Kholiavko, N., et al. (2020). "Budget Management Systems in Modern Financial
Environments," Journal of Economic Studies, Vol. 48, Issue 3.
o Lee, R. D., Johnson, R. W., & Joyce, P. G. (2019). Public Budgeting Systems. Jones &
Bartlett Learning.
3. Whitepapers:
o "Improving Budgetary Performance through Automation," Gartner Research, 2021.
(Discusses how software solutions enhance budget management.)

Online Resources
1. Software Development Guides:
o IBM’s Budget Management Framework: IBM Documentation o Oracle’s Financial
Planning and Budgeting Tool Overview: Oracle Docs
2. Financial Management Resources:
o Mint and YNAB (You Need A Budget) systems: Case studies and best practices.
o Open-source projects on GitHub (search for budget management systems): Examples of
real-world implementations.

Standards and Guidelines

1. Financial Standards:
o ISO 22301 (Business Continuity Management Systems): For ensuring financial
operations' stability.
o GAAP (Generally Accepted Accounting Principles): Core principles for financial
reporting and budgeting.
2. IT and System Design Standards:
o IEEE 830-1998: Standard for Software Requirements Specifications.
o OWASP (Open Web Application Security Project): Guidelines for securing financial
systems.
 Tools for Implementation
1. Development Tools:
o Frontend: React, Angular. o
Backend: Node.js, Django.
o Database: PostgreSQL,
MongoDB.
2. Testing Tools:
o Selenium for UI automation. o
JMeter for performance
testing.
3. APIs and Integrations:
o Open Banking APIs: Plaid,
Yodlee. o Payment Gateways:
Stripe, PayPal.

General Best Practices

1. Government Budgeting Resources:

o US Government Accountability Office (GAO): Guides on budget management
for public projects. o IMF’s Financial Management Toolkit: IMF Resources.
2. Case Studies:
o Case studies from Deloitte and PwC on enterprise financial planning tools. o
Success stories of companies using automated budgeting solutions like SAP or
Microsoft Dynamics.

Appendix A: Screenshots of the Budget Management System

1. Login and Registration Page

Screenshot of the user login interface showing:

o Email/username and password fields. o

"Forgot Password?" and "Sign Up" options.

3. Dashboard

Overview of the user's financial summary:

o Pie chart showing expense categories.
o Bar graph displaying monthly spending trends. o Available budget
balance.

4. Budget Creation Interface

 Input fields for:
o Budget name (e.g., "Groceries").
o Allocated amount.
o Duration (weekly, monthly, yearly).

5. Expense Tracking Interface

A list view of recent transactions with:

o Date, category, and amount fields. o
Options to add/edit/delete expenses.