TravelCopilot

Introduction

A travel recommendation system that uses matrix factorization and lightGCN to generate personalized travel plans based on user preferences and historical data from the Yelp Academic dataset. The system employs three different recommendation models for attractions, hotels, and restaurants, respectively, to optimize the travel plan for efficiency and accuracy. The system also features a user-friendly graphical interface developed using PyWebIO. This repository contains the source code and documentation for Travel Copilot.

Display of results

Click here to view the results of a full run.

Document description

.
├── Readme.md															
# this file 
├── code																	
# code directory
│   ├── Data Processing and Analysis										
# data processing and analysis
│   │   ├── Data_Load_And_Cleaning_Business.ipynb						
# data cleaning and processing of business data
│   │   ├── Data_Load_And_Cleaning_Review.ipynb							
# data cleaning and processing of review data
│   │   ├── loading-yelp-json-data-and-visualization.ipynb				
# data loading and visualization
│   │   ├── nohup.out													
# log file for a training process
│   │   ├── mkyelp2018.py												
# data processing for Light-GCN												
│   │   └── plot.ipynb													
# data visualization for log file
│   ├── Procedure.py														
# test and train function
│   ├── TravelCopilot.py													
# TravelCopilot system entrance
│   ├── checkpoints														
# model checkpoints directory
│   │   └── lgn-yelp2018-3-64.pth.tar									
│   ├── dataloader.py														
# dataloader in yelp		
│   ├── filter.py															
# filter function for business data
│   ├── generate_guide.py													
# generate guide function
│   ├── idIndex.py															
# idIndex function for business ID and business number in Light-GCN
│   ├── model.py															
# model definition
│   ├── parse.py															
# parse function 
│   ├── rcmdweb.py															
# web UI and user interaction
│   ├── recommendation.py													
# recommendation function for tour, hotel and restaurant
│   ├── register.py															
# register function 
│   ├── train.py															
# train function
│   ├── utils.py															
# utils function
│   └── world.py															
# world function 
├── data																
# data directory for Light-GCN	
│   └── yelp2018															
# data directory for yelp
│       ├── item_list.txt													
# business Id and business number
│       ├── s_pre_adj_mat.npz												
# pre-processed adjacency matrix
│       ├── test.txt														
# test data
│       ├── train.txt														
# train data
│       └── user_list.txt													
# user Id and user number				
├── doc																	
# document directory
│   ├── TravelCopilot.pdf												
# results of a full run
│   ├── TravelCopilot.pptx													
# ppt of TravelCopilot				
│   └── report.pdf															
# report of TravelCopilot		
├── output_csv															
# data directory for SVD		
│   ├── business_PA_Philly_clean.csv										
# business data after cleaning	
│   ├── business_PA_Philly_clean_tour.csv									
# business data after cleaning and filtering	
│   ├── review_PA_Philly_clean.csv											
# review data after cleaning
│   └── review_PA_Philly_clean_tour.csv										
# review data after cleaning and filtering
└── yelp_dataset														
# original yelp2023 data directory
    ├── Dataset_User_Agreement.pdf										    
    ├── yelp_academic_dataset_business.json	
    ├── yelp_academic_dataset_checkin.json	
    ├── yelp_academic_dataset_review.json	
    ├── yelp_academic_dataset_tip.json	
    └── yelp_academic_dataset_user.json	

Train your own model

train LightGCN on Yelp dataset:

• command

 cd code && python train.py --decay=1e-4 --lr=0.001 --layer=3 --seed=2020 --dataset="Yelp2018" --topks="[20]" --recdim=64

Run TravelCopilot

• command

cd code && TravelCopilot.py	

Reference

https://github.com/gusye1234/LightGCN-PyTorch