search

byunsy / financial-forecasting

Uses a combination of LSTM (Long Short-term Memory) and CNN (Convolutional Neural Network) to accurately predict a stock's closing price based on a time-window of 20 business days.
5 stars 3 forks source link

readme

financial-forecasting

image

Description

Financial-forecasting is a jupyter notebook that effectively utilizes both LSTM (Long Short-term Memory) and CNN (Convolutional Neural Network) to accurately predict a stock's closing price based on a time-window of 20 business days. Firstly, I used the FinanceDataReader package to conveniently scrape specific stock market data and constructed a neural network model that combined a mixture of LSTM and CNN models. I then used over 6,000 data points (4,800 for training and 1,200 for testing) to create a financial forecasting model given a specific stock item. To process large sets of data points efficiently, I decided to use Google Colab, which allowed me to utilize its GPUs and memory space.

The general process for the notebook was:

  1. Import all necessary packages
  2. Attain stock market dataset from FinanceDataReader
  3. Data Exploration and Visualizations
  4. Data Preprocessing
  5. Creating a LSTM/CNN model
  6. Compiling the LSTM/CNN model
  7. Training the CNN model
  8. Model Prediction

Usage

Every detailed step was recorded in the notebook and can easily be followed there.

Figures and Charts

Data Visualization

Model Prediction