Improving Story Points Estimation Using Ensemble Machine Learning

This project aims to develop and implement a machine learning-based model for story point estimation using the ensemble stacking technique, combining RoBERTa and BiLSTM models.

Table of Contents

• Setup Environment
• Preprocessing
• RoBERTa Model
• BiLSTM Model
• Stacking
• Usage

Setup Environment

1. Install Anaconda
2. Create the environment:

   conda env create -f ml_sp_env.yml

3. Execute additional requirements in info.ipynb

Preprocessing

Run the preprocessing steps in preprocess.ipynb to prepare the datasets for training and evaluation.

RoBERTa Model

The RoBERTa model is implemented in roberta-base.ipynb. Follow the notebook to train and evaluate the RoBERTa model.

BiLSTM Model

The BiLSTM model is implemented in bilstm-base.ipynb. Follow the notebook to train and evaluate the BiLSTM model.

Stacking

The stacking approach is implemented in stacking.ipynb. This notebook combines the predictions from the RoBERTa and BiLSTM models to improve the overall performance.

Usage

1. Preprocess the data using preprocess.ipynb.
2. Train and evaluate the RoBERTa model using roberta-base.ipynb.
3. Train and evaluate the BiLSTM model using bilstm-base.ipynb.
4. Perform stacking using stacking.ipynb.