genetically_optimized_graph_RS

This is the source code for our paper Graph-based Recommendation for Sparse and Heterogeneous User Interactions.

Requirements

• Pandas
• Pandasql
• NumPy
• NetworkX
• Operator
• Scikit-learn
• Joblib
• Multiprocessing
• Recommenders
• RecBole
• PyTorch
• SciPy
• PyGAD
• Pickle
• Json
• Statsmodels
• Maatplotlib

Datasets

We use the educational social network dataset at https://github.com/carmignanivittorio/ai_denmark_data and the insurance dataset at https://github.com/simonebbruun/cross-sessions_RS.

Usage

Educational social network:

1. Split the data into training and test set using
   1_data_split.py
2. Tune the hyperparameters of the models using
   2_graph_based_model_tune.py
   2_KNN_tune.py
   2_NeuMF_tune.py
   2_NGCF_tune.py
   2_SVD_tune.py
3. Optimize the weights of the graph-based models with genetic algorithm using
   3_directed_graph_weight_optimization.py
   3_undirected_graph_weight_optimization.py
4. Evaluate the models over the test set using
   4_KNN_evaluation.py
   4_KNN_varying_thresholds.py
   4_most_popular_evaluation.py
   4_most_popular_varying thresholds.py
   4_NeuMF_evaluation.py
   4_NeuMF_varying_thresholds.py
   4_NGCF_evaluation_and_varying_thresholds.py
   4_SVD_evaluation.py
   4_SVD_varying_thresholds.py
   4_undirected_graph_evaluation.py
   4_undirected_graph_varying_thresholds.py
   4_uniform_and_directed_graph_evaluation.py
   4_uniform_and_directed_graph_varying_thresholds.py
5. Plot evaluation measures for varying thresholds and test for statistical significans using
   5_varying_thresholds_plot.py
   5_statistical_significans.py

Insurance:

1. Preprocess the data using
   1_data_preprocessing.py
2. Tune the hyperparameters of the models using
   2_graph_based_model_tuning.py
   2_KNN_tuning.py
   2_NeuMF_tuning.py
   2_NGCF_tuning.py
   2_SVD_tuning.py
3. Optimize the weights of the graph-based models with genetic algorithm using
   3_directed_graph_weight_optimization.py
   3_undirected_graph_weight_optimization.py
4. Evaluate the models over the test set using
   4_KNN_evaluation.py
   4_most_popular_evaluation.py
   4_NeuMF_evaluation.py
   4_NGCF_evaluation.py
   4_SVD_evaluation.py
   4_undirected_graph_evaluation.py
   4_uniform_and_directed_graph_evaluation.py
5. Plot evaluation measures for varying thresholds and test for statistical significans using
   5_varying_thresholds_plot.py
   5_statistical_significans.py