GenRS 0.0.5

A Generative learning-based Framework for Recommendation System algorithms

GenRS

A generative learning-based Framework for Recommendation Systems algorithms

Software requirements:

• Python 3.6 or higher
• Tensorflow 1.15 GPU
• Numpy 1.17
• Scipy 1.3
• Pandas 0.25
• Bottleneck 1.3

Algorithms list available:

• CFGAN
• IRGAN
• EASE
• VAE
• DAE

TO DO:

• Check which dataset you want to use from here
• Download and extract the preferred

Set the framework configuration

• Download cfg.JSON file from https://github.com/cedo1995/GenRS/tree/master/Cfg

• Check if path contains the path to your chosen dataset file

• Check separator (sep) used in selected dataset and update it if necessary

• Check algos you want to compute respecting the list of string lowercase format as predefined

• Define the number of users to use as validation and test set through heldout_us_val and heldout_us_test param

• Check metrics you want to compute from: ["precision@k", "recall@k", "ndcg@k", "ap@k", "auc"]

Set the algorithms configuration

• Download {alg}_cfg.JSON where {alg} correspond to name of algos previously set in algos parameter in cfg.JSON file

Import RecSys module by:

from GenRS.Main.rec_sys import RecSys

Define the path to cfg.JSON file and {alg}_cfg.JSON files

Execute

RecSys(path_frm_cfg, list_algs_cfg_path)

Results will be into console.log.txt file