Time series

Summary: Exploring different machine learning algorithms to predict stock prices in an interval 1-10 days, on historical data.

Keywords: Machine Learning, Scikit-learn, TensorFlow

Repository: timeseries_forecasting_ml

Date: 3 April 2025 (work in progress)

Stock price predictions with different machine learning algorithms

As a first working example I take the Apple stock price from 01-01-2020 to 01-01-2024. The models are trained on these data minus the last 10 days, and then predict the value for these days. I use the closing price and train the models with lag features, the past 5 closing prices with respect to the target value. The models are:

Linear Regression (LR): fits only the 5 lag features
Neural Network (NN): feedforward NN with two hidden layers of 20 fully connected neurons and ReLu activation function
Gated Recurrent Unit (GRU): 2 layers of 20 units
Long Short-Term Memory (LSTM): 2 layers of 20 cells

For the three different neural networks, in this experiment, I used an Adam optimizer with learning rate 0.001, mean squared error (MSE) as the loss function, 50 epochs for training, and batches of size 64 (that is training data points process simultaneously). Based on the MSE of the predictions the ranking is the following (best to worst): LR, GRU, NN, LSTM.

Updates on more experiments and results are on the way.