An Explainable AI-Driven Requirements Engineering Approach: Mining app review by Analyzing Low-Rating Feedback

XARM: Explainable App Reviews Mining Approach This repository contains datasets and code for conducting experiments on review mining with a focus on explainable AI techniques. The work explores preprocessing, sampling methods, and deep learning models for extracting insights from textual review data. Directory Structure Data Files Amazone_Dataset.csv: Contains review data for experiments. annotatted_dataset.csv: Annotated dataset used in various models. Code Files Running the Experiments Preprocessing and Sampling and Training ML learning Models Oversampling_and_Under_samplimpling_countvatrizer.ipynb: Demonstrates cleaning text data and balancing datasets using oversampling/undersampling techniques with CountVectorizer. Oversampling_and_Under_samplimpling_tfiddf_multiclassification.ipynb: Similar to the above but utilizes TF-IDF for text representation. Preprocessing and Sampling and Training Deep Learning Models CNN Modell.ipynb: Uses Convolutional Neural Networks for review mining. GRU Model.ipynb: Employs a GRU-based architecture for text classification. LSTM, BILSTM_models.ipynb: Explores LSTM and BiLSTM models for review classification. RNN_MODEL.ipynb: Implements the RNN model for review classification. ROC BILSTM_OVER_UNDER_SAMPLING.ipynb: Focuses on evaluating the BiLSTM model's performance with balanced datasets. BiGRU_MODEL.ipynb:Implements the BiGRU model for review classification. SHAP and LIME Experiment Files using ML and DL learning CNN LIME.pdf: Analysis of CNN model using LIME (Local Interpretable Model-agnostic Explanations). CNN_SHAP.pdf: SHAP (SHapley Additive exPlanations) analysis applied to CNN model. MLP_LIME.pdf: Analysis of the MLP model using LIME. shap_mlp.pdf: Analysis of the MLP model using SHAP. Requirements Python 3.x Jupyter Notebook or JupyterLab Libraries: pandas, numpy, scikit-learn, keras, tensorflow Evaluation Use the associated PDF reports for insights on model interpretability using LIME and SHAP.