Skip to content

Feature Engineering and Predictive Modeling for Financial Time Series Data

Notifications You must be signed in to change notification settings

lmedeiro/fe_and_pm_for_financial_time_series

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Repo Description

This repo contains supporting files and report for a research project done for Aalto University. All material on this repo is free to be explored for your own use. If there are any questions, please contact me. If you wish to cite or use it for any further purposes, please contact me beforehand.

For the report, you may refer to this pdf . Each jupyter notebook has its own model and purpose, therefore you may visit it and run the code. Models and machine learning related procedures are implemented with Pytorch, Skorch, and Scikit-learn. The main models used for this project were: LSTM RNNs, CNNs, and RFs.

The main dataset used for this work is in this pickle file . It contains the structure that is known by the programs already. A few other pickle files are provided as well. You may test with those if desired.

Research Abstract

Feature Engineering and Predictive Modeling for Financial Time Series Data

This paper presents a series of results achieved by attempting to predict the closing price of a financial market time series. The data source was Yahoo Finance, containing 19 years worth of daily stock information, and an ETF tracking the Nasdaq 100 index (QQQ) was chosen as the specific dataset to predict. A set of features for financial time series data was engineered and five (5) models were explored: Random Forest, LeNet, custom LeNet, ResNet, and a LSTM. The goal was to use to Deep Neural Network models, while engineering features that allow reliable prediction with only a single day worth of information. The best performing models managed to achieve accuracy ranges between 83-85%, showing that the features engineered have predictive value.

About

Feature Engineering and Predictive Modeling for Financial Time Series Data

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published