StockPredictionML Presentation
- 1. AAPL STOCK PREDICTION MODEL DUSTIN JASMIN
- 2. GOAL • Develop a Machine Learning model that uses: • Relative Strength Index (RSI) • Moving Average Convergence Divergence (MACD) • To predict future % returns on AAPL shares • Build trading bot that uses model to make buy and sell orders based on live market data
- 3. PROCESS • Imported many libraries • Notables: • Technical Analysis (TA) • NumPy, Pandas • Alpaca API • CometML • Pytorch, Keras, Tensorflow
- 4. PROCESS, CONTINUED • Used Alpaca to import AAPL stock data (2016-2018): • Each data point was in 4-hour intervals • Used RSI, MACD and Closing Share Price for each data point • Converted each dataset to NumPy arrays • Developed Multi-Layered Perceptron • Feed forward neural network • Consists of three types of layers—the input layer, output layer and hidden layer • The input layer receives the input signal to be processed. • Output layer performs prediction and classification • An arbitrary number of hidden layers that are placed in between the input and output layer are the true computational engine of the MLP.
- 5. CODE, PRE-PROCESSING • Added data to DataFrames • Close series = 2 years of AAPL closing price • Added close series to RSI() and MACD() object • Converted to NumPy array
- 6. CODE, NORMALIZE • Set data to between 0 and 1 for easier computation • Used unsqueeze() to format data for training
- 7. CODE, MODEL & PREDICTION
- 8. PERFORMANCE • Model Learning Loss has sharp decline in beginning • Steady decline afterwards • Epoch duration (time to cycle through dataset) decreases over time
- 9. SUMMARY • Did not get to develop trading bot • Personal Assessment: • Based off loss data, model seemed to performed decently • Project made me research/learn more about ML • Data processing is toughest part in ML