Shanlax

International Journal of Arts, Science and Humanities

OPEN ACCESS Drug Recommendation System

Volume: 11
Using Machine Learing
Special Issue: 1 Priyanka V.G
Department of Computer Applications
Raja RajeswariCollege of Engineering
Month: July
Pushpalatha G
Year: 2023 Department of Computer Applications
Raja RajeswariCollege of Engineering
E-ISSN: 2582-0397
Abstract
P-ISSN: 2321-788X Machine learning’s value has soared in diverse applications, driving automation in
creative work. We present a medication recommendation system to lighten experts’
workload. Our method suggests medications effectively. COVID-19 has strained
Impact Factor: 3.025 clinical resources, leading to scarcity of experts, equipment, and medications.
People resort to self-medication without guidance, worsening the health situation.
Received: 08.05.2023 Bow, TF-IDF, Word2Vec, and Manual Feature are the tools we employ. for
emotion prediction. By employing categorization algorithms, we analyse and select
appropriate medications. Accuracy, accuracy, remembrance, f1score, and AUC
Accepted:13.06.2023 score are all important metrics. evaluate predictive sentiments. Results favour
LinearSVC with TF-IDF.Vectorization,works superior to every other model, with
Published: 01.07.2023 an accuracy rate of 93%.
Keywords: Medicine, Machine Learning, Patients.
Citation:
Priyanka, VG, and G. Introduction
Pushpalatha. “Drug The rapid increase in coronavirus cases has led to a doctor shortage,
Recommendation especially those in rural areas. it difficult to address the shortage
System Using Machine quickly. In this challenging period, activating telemedicine is crucial.
Learing.” Shanlax Medication errors are prevalent, causing harm to thousands of people
International Journal each year. Due to limited information, doctors often prescribe the
of Arts, Science and wrong medication, resulting in a 40% error rate. Selecting the best
Humanities, vol. 11, prescription is challenging as new research and medications emerge
no. S1, 2023, regularly.
pp. 231–35. The importance of item reviews has grown with the expansion of
the internet and online commerce. People rely on to write reviews
DOI: informed decisions about purchases. However, the healthcare area
https://doi.org/10.34293/ has gotten fewer spotlights nowadays. this area. More individuals are
sijash.v11iS1-July.6344 seeking online diagnoses, making system essential.
To aid doctors and patients in understanding medications
for specific medical problems, now a days.a framework for
recommendations is necessary. This system uses the engineering of
features and sentiment analysis to suggest drugs based on patient
evaluations. Sentiment analysis identifies emotional information from
language, while feature engineering enhances model performance by
adding new features.
http://www.shanlaxjournals.com 231
 Shanlax
International Journal of Arts, Science and Humanities

Literature Survey
1. Machine Learning-Based Sentiment Analysis of Drug Reviews for Drug Rating and
Recommendation
According toMd. Deloar Hossain,Users may make educated judgements with the aid of a
recommendation system based on sentiment analysis. from complex knowledge. However, applying
sentiment analysis to healthcare and medication experiences is challenging. This study introduces
a medication recommendation system that uses sentiment analysis of medication reviews. The
objective is to help patients choose medications with more knowledge. The approach analyses
sentiment in reviews of medical products, examines the sentiment polarity of reviews of drugs and
patient conditions, and determines the value of reviews to users. We include pertinent medication
properties into the algorithm for making recommendations. In addition to the Hybrid model for
recommendation, the system is assessed using the Decision Tree, K-Nearest Neighbour, and
Linear Support Vector Classifier algorithms for rating creation. Fine-tuning the settings improves
performance. Classifier for linear support vectors is chosen for rating creation, balancing accuracy,
efficiency, and scalability.
2. Using implicit feedback and crossing recommendations, a system for prescribing drugs
for epilepsy
According to ChunChen; LuZhang; Xiaopeng Fan; Yang Wang; Chengzhong Xu.
Epilepsy is a chronic neurological condition characterized by unexpected brain electrical
disturbances. Treating epilepsy is challenging, but around 70% of seizures can be controlled with
drugs. Electronic Health Records (EHRs) of epileptic patients offer essential details for customised
treatment. This study analyses real medical cases to propose an Implicit Feedback and Crossing
prescription (IFCR) system for medication selection. The approach considers the medical history
of epileptic patients to identify connections between symptoms and medications. Compared to an
ANN-based system, our solution achieves higher recall rates, with up to a 30% increase. Overall,
IFCR outperforms ANN in terms of performance. Combining IFCR and ANN in an ensemble
model leverages the unique strengths of each algorithm.

Existing Methodology
“Deep Adverse: Deep Learning for Adverse Drug Event Prediction” by Luo et al. (2018): This
study utilizes deep learning techniques, specifically recurrent neural networks (RNNs), to predict
adverse drug events. The methodology involves training RNN models on patient drug usage data
and clinical features to accurately identify potential adverse reactions.
“Personalized Drug Recommendation System Based on Deep Learning and Patient Similarity”
by Sarker et al. (2019): This research introduces a personalized drug recommendation system
that combines deep learning techniques and patient similarity. The methodology involves using
convolutional neural networks (CNNs) to extract features from patient data, and then leveraging
patient similarity measures to recommend drugs tailored to an individual’s medical profile.
“Drug Recommendation System Based on Drug Similarity and Disease Similarity” by Wang
et al. (2020): The authors propose a drug recommendation system that considers drug similarity
and disease similarity. The methodology involves constructing a drug-drug similarity matrix and
a disease-disease similarity matrix based on data integration and similarity calculation techniques.
Recommendations are then generated by combining these matrices to find similar drugs for a given
disease.
“Drug Recommendation Using Machine Learning Techniques” by Pablo Tobón et al. (2019):
This research paper proposes a drug recommendation system that combines collaborative filtering
and content-based filtering approaches. The authors utilize patient demographics, medical history,
232 RajaRajeswari College of Engineering, Bangalore
 Shanlax
International Journal of Arts, Science and Humanities

and drug properties to generate personalized drug recommendations using machine learning
algorithms such as decision trees and random forests.
“Enhancing Medication Guidance through Recurrent Neural Networks: A Paradigm for Drug
Recommendation” by Juan Du et al.: This paper presents a drug recommendation system based
on recurrent neural networks (RNNs). The authors use patient electronic health records and drug
information to train an RNN model that captures temporal dependencies and sequences in patient
data. The model then predicts suitable drugs for individual patients.

Proposed System
An item is suggested to the user via a common system called a recommender framework based
on their benefit and necessity. These frameworks make use of consumer surveys to assess client
sentiment and provide recommendations for their particular need. The medicine recommender
system uses sentiment analysis and feature engineering to conditionally recommend drugs
depending on patient input. separation and extraction of emotional content from words, such
as judgements and attitudes, is known as sentiment analysis [7]. The technique of “feathering
engineering,” nonetheless, involves expanding models’ capability by adding new features to the
ones that currently exist. The Django web framework will be utilised in this project, and the user
will enter the prescription pill in accordance with the drug advice we will provide based on the
prediction outcome.

Figure 1 Proposed Architecture

Machine learning may be used to create a drug recommendation system that will help doctors
choose the best treatments for their patients. The following process is suggested for creating such
a system:
Data Collection: assemble an extensive dataset with details on individuals’ health problems,
histories of treatments, the success of their medications, any negative side effects, and any
additional pertinent factors. Electronic health records, clinical studies, medical literature, and other
trustworthy sources are good places to find this information.
Data Preprocessing: Clean the collected data by removing any inconsistencies, errors, or
missing values. Perform data normalization and feature engineering to prepare the dataset for
machine learning algorithms.
Feature Selection: Decide which aspects of the drug recommendation process are most
important. Statistical analysis, domain knowledge, or feature selection techniques can all be used
to do this. Patient demographics, medical history, symptoms, genetic data, and lab test findings can
all be significant factors.

http://www.shanlaxjournals.com 233
 Shanlax
International Journal of Arts, Science and Humanities

Result
Several performance metrics, including as accuracy, precision, recall, F1 score, and AUC-ROC,
may be used to assess drug recommendation systems. rigorous confirmation techniques such as
dataset splitting, cross-validation, and user studies are crucial to ensure meaningful results. Expert
recommendations and feedback from healthcare professionals can be used for comparison and
assessment. Developing a drug recommendation system requires collaboration between domain
experts, data scientists, and healthcare professionals. Continuous monitoring, refinement, and
updating with new data are essential to improve accuracy and relevance over time.

Figure 2 Registration Page

Figure 2 shows the register page of drug recommender system.

Figure 3 Drugs Available Per Conditions in the Patients

Figure 3 shows the most drugs available per conditions in the patients.

Conclusion
Drug recommendation systems using machine learning have emerged as a promising approach
to assist healthcare professionals in making personalized and effective drug recommendations. The
234 RajaRajeswari College of Engineering, Bangalore
 Shanlax
International Journal of Arts, Science and Humanities

methodologies employed in these systems vary but often involve collaborative filtering, content-
based filtering, knowledge-based approaches, hybrid models, deep learning techniques, and graph-
based methods.
Machine learning algorithms analyse large-scale patient data, electronic health records, drug
properties, and domain knowledge to generate accurate and tailored drug recommendations. These
systems leverage patient similarities, drug attributes, historical data, and complex patterns to
provide personalized suggestions, considering factors such as drug efficacy, safety, and patient-
specific characteristics.
The use of machine learning in drug recommendation systems has the potential to enhance patient
outcomes, improve treatment adherence, and reduce adverse drug events. However, challenges
such as data quality, interpretability, scalability, and integration with clinical workflows need to be
addressed for widespread adoption.
Further research and development are required to refine and validate these systems, considering
real-world evidence, heterogeneous data sources, and the unique needs of diverse patient
populations. By continuously advancing the methodologies and incorporating new data sources,
machine learning-based drug recommendation systems hold great promise for revolutionizing
personalized medicine and optimizing patient care.

References
1. Zhang, Y., Zhu, Y., Hao, J., & Huang, J. (2017). A survey on recommendation algorithms
for clinical decision support. IEEE/ACM Transactions on Computational Biology and
Bioinformatics, 14(3), 710-737
2. Cheng, F., Desai, R. J., Handy, D. E., & Wang, R. (2018). The integration of network analysis
and population-based validation enables the prediction and verification of in silico drug
repurposing stratergies, Link: https://doi.org/10.1038/s41467-018-02801-9
3. Chen, B., Ma, L., Paik, H., Sirota, M., Wei, W., & Chua, K. (2019). The identification of
therapeutic targets is facilitated by the correlation between drug efficacy and the reversal of
cancer gene expressionpatterns, Link: https://doi.org/10.1038/s41467-019-11035-5
4. Bhattacharyya, R., &Basu, S. (2018). India Inc looks to deal with rising stress in employees.
Retrieved from ‘The Economic Times’
5. OSMI Mental Health in Tech Survey Dataset, 2017 from Kaggle
6. Van den Broeck, J., Cunningham, S. A., Eeckels, R., & Herbst, K. (2005). Data cleaning:
detecting, diagnosing, and editing data abnormalities. PLoS medicine, 2(10), e267.
7. Shwetha, S, Sahil, A, Anant Ku mar J, (2017) Predictive analysis using classification
techniques in healthcare domain, International Journal of Linguistics & Computing Research,
ISSN: 2456-8848, Vol. I, Issue. I, June-2017
8. Tomar, D., & Agarwal, S. (2013). A survey on Data Mining approaches for Healthcare.
International Journal of Bio-Science and Bio-Technology, 5(5), 241-266.
9. Gender and Stress. (n.d.). Retrieved from APA press release 2010.
10. “Drug Recommender Systems: A Framework for Representation and Evaluation” by Alan R.
Misselbrook and Benjamin Gyori (2017) - The authors present a framework for drug

http://www.shanlaxjournals.com 235