Time Series Experiments
Project Motivation -
Financial time series is fundamental to the field, and predicting stock price has been a challenging problem to solve. I am going to attempt to use CRISP-DM approach to understand the patterns in data and answer some practical questions about investment decisions investors need to make.
Questions asked
- What is the level of risk — nature of the distribution of returns?
- What is the contribution of trend, seasonality, and noise (risk)?
- Which features are co-related most with a price?
- Does price have some memory of past — autoregressive structure?
- Can we predict stock price directional movement or price itself?
The project explains how the CRISP-DM process can be applied to financial time series data. CRISP-DM is the cross-industry process for data mining. It is a structured approach to planning a data mining project. Code along with the medium article (https://disruptivenext.medium.com/crisp-dm-for-financial-time-series-b4e01fcb4e8b) explains how meaningful patterns can be detected in Financial time series data. Project further proposes key questions that can help the investor to make investment decisions.
Requirements
- TA_Lib==0.4.19
- scikit_plot==0.3.7
- yfinance==0.1.54
- scipy==1.4.1
- numpy==1.18.1
- statsmodels==0.11.0
- pandas==1.0.1
- seaborn==0.10.0
- matplotlib==3.1.3
- scikit_learn==0.24.1
- scikitplot==0.1.1
- talib==0.1.1
Files in the Repository
- DataAgreegator.ipynb - Program to download data from yahoo finance
- EDA.ipynb - Exploratory data analysis for financial time series
- Model.ipynb - Simple linear regression model for financial time series
Data source used Basic data is downloaded from yfainance then augmented with additional features using TA-Lib.
Results Summary has been elaboarted in https://disruptivenext.medium.com/crisp-dm-for-financial-time-series-b4e01fcb4e8b.
Acknowledgment All of the content is purely based on my work experience in the financial field. API documentation from respective libraries is used to understand how to use the API.