Skip to content

Stock target prediction - Predicting Money movenent in stock by using Ensemble Methods

Notifications You must be signed in to change notification settings

AdhvikVithun/Stock-Target-Movement

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 

Repository files navigation

Predictive Financial Analysis with Ensemble Models

** To view ouput , ppt and conference paper go to Ouputes file **

Overview: This repository contains a comprehensive Python program for predictive financial analysis, focusing on stock price forecasting using a diverse set of machine learning models. The project employs time series analysis, deep learning, and ensemble techniques to enhance the accuracy of stock price predictions.

Program Flow Program flow 2

Table of Contents:

Introduction

Data Collection

Feature Engineering

Target Variable Setup

Exploratory Data Analysis

Scaling and Preprocessing

LSTM-based Autoencoder

XGBoost Regressor

LSTM Model

Ensemble Approach

Results and Evaluation

Conclusion

Introduction: The program utilizes various machine learning models, including LSTM-based autoencoders, XGBoost regressors, and Support Vector Machines (SVM), to predict future stock price movements. The ensemble approach combines these models to achieve a more robust and accurate forecasting mechanism.

Data Collection: Historical stock price data for the company 'META' is collected using the Yahoo Finance API. The data is stored in a CSV file for further analysis.

Feature Engineering: Technical indicators such as Relative Strength Index (RSI) and Exponential Moving Averages (EMAs) are computed and added to the dataset for better model performance.

Target Variable Setup: The target variable is defined as the difference between the adjusted closing price and the opening price of the stock. A binary classification ('TargetClass') is also created based on the direction of this difference.

Exploratory Data Analysis: Visualizations are provided to explore the distribution and trends of the target variable, offering insights into the dataset.

Scaling and Preprocessing: Data is scaled using Min-Max scaling, and the dataset is split using TimeSeriesSplit for training and testing purposes.

LSTM-based Autoencoder: A deep learning model with LSTM layers is implemented to capture temporal patterns in the data, enhancing feature representations.

XGBoost Regressor: An XGBoost regressor is trained using the encoded representations from the autoencoder, contributing to the predictive ensemble.

LSTM AE + XG

LSTM Model: A standalone LSTM model is created to further exploit temporal dependencies in the dataset.

Ensemble Approach: The final predictions are generated by combining the outputs of the XGBoost regressor, SVM regressor, and the LSTM model, providing a more robust forecast.

ensemble 1

Results and Evaluation: Evaluation metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) are employed to assess the accuracy of individual models and the ensemble approach.

Conclusion: The program showcases a holistic approach to stock price prediction, leveraging the strengths of various machine learning techniques. The ensemble model demonstrates improved accuracy, providing a valuable tool for financial analysts and investors.

Note: The provided code and analysis are for educational and experimental purposes. The predictive accuracy may vary based on data quality, market conditions, and other factors.

Just tried this model: Just Tried this Model

About

Stock target prediction - Predicting Money movenent in stock by using Ensemble Methods

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages