This project is a movie recommendation system that uses Collaborative Filtering to predict user preferences. The system is built in C++.
The recommendation algorithm operates on user-item rating data, represented as pairs in vectors. The system can identify the top users and movies by their watch count.
The recommendation system uses two files: train.csv and test.csv. The train.csv file is used for creating the necessary data structures, while the test.csv file is used to evaluate its performance. The final predictions are written to submission.csv.
The project can be run by compiling and executing the main.cpp file. Make sure the train.csv and test.csv files are in the same directory as the main.cpp file.
Please note that the project also includes createAllList.h and ratePredictor.h header files which should also be in the same directory.
g++ main.cpp -o main
./main
This will run the recommendation system and write the predictions to submission.csv.
The system includes performance timing, outputting the time taken to find the top 10 users and movies as well as the total running time of the algorithm.
The average time for it to find Top 10 Users/Movies after creating the necessary data structures is 1.3 milliseconds for 35200 users and 7331 films.
The average time for the algorithm to make 5,000 predictions, given ~1.1 million training data, is 125 seconds.
Performance measurements have been made with 8GB/256GB M1 Macbook Air.
- C++ Standard Library
<chrono>for performance timing<fstream>for file input/output<vector>and<pair>for data structure management<iostream>for console input/outputcreateAllList.hfor creating the necessary data structuresratePredictor.hfor making predictions with processed data
Please refer to Project Details pdf file for a more detailed explanation of the system's workings.