chiggy1997 / Stock-Price-Prediction

Stock Price Prediction

Stock-Price-Prediction

• In this project, I trained a Ridge regression model and deep neural network model to predict future stock prices.
• The AI model will be trained using historical stock price data along with volume of transactions.

1. Overview :

a. Goal :

• Based on volume and price data , can we predict stock prices?
• Why?
• Because while investing other peoples money we need to be really careful and we need real data to support our decisions rather than intuition

b. Impact :

• Optimization and revamp which improved the accuracy by 12%

c. Challenges Faced :

• Reshaping data while feeding into the model
• Optimization which needed multiple level of changes and multiple runs plus going through numerous research papers.

d. Interesting findings :

• With only a few layers we can get good accuracy.

2. Code and resource used :

• Data manipulation: Pandas
• Visualisation : Plotly and matplotlib
• Regression : Sklearn
• Metrics : R2score
• Deep Learning : Tensorflow, keras

3. About the data :

4. Data Preprocessing :

• Took data of all stocks and combined them in a single dataframe containing only the strike price and close price
• Time based splitting
• Normalization

5. Visualisation :

• Histogram of returns data in matplotlib.py and plotly express to understand its distribution
• Volatility vs time plot to observe the change in volatily over time.
• Price vs time to observe the stock returns over time

6. Modelling :

• Ridge regression with accuracy as metrics.
• LSTM model with following layers
• Optimizer used was ADAM
• Loss taken was categorical mean squared error and metrics taken as accuracy.