louis925/kaggle-santa-gift-matching-challenge

A Monte Carlo and max-flow approach

Kaggle - Santa Gift Matching Challenge

Monte Carlo and Min-Cost Max-Flow Approach

Solutions to the 2018 Kaggle optimization competition - Santa Gift Matching Challenge

Authors: Louis Yang and Sky Huang

Methods: Monte Carlo, Min-Cost Max-Flow (MCMF), and Mixed-Integer Programming (MIP).

Requirement

Max-Flow and MIP approaches require Google Ortools.

Installation on Microsoft Visual Studio 2017

1. Download and unzip OR-Tools for C++ (Visual Studio 2017)
2. Create an empty C++ Visual Studio solution for one of the folder.
3. Change configuration to "Release" and platform to "x64".
4. In Solution Explorer, right click on the project > Properties.
5. In VC++ Directories, ([or-tools] is the directory where you unpacked the or-tools archive.)
   1. add "[or-tools]\include" to "Include Directories".
   2. add "[or-tools]\lib" to "Library Directories".
6. In Linker > Input, add "ortools.lib" to "Additional Dependencies".

Descriptions

• input - Download and upzip "gift_goodkids_v2.csv", "child_wishlist_v2.csv", and "sample_submission_random_v2.csv" to here from Kaggle.
• max_flow_mc - Monte Carlo with Min-Cost Max-Flow approach
• max_flow_merge_mc - Improved Monte Carlo with Min-Cost Max-Flow approach which merges groups
• max_flow_relax_all - Solve the complete relaxed problem as Min-Cost Max-Flow problem
• max_flow_search
• mc_swap - Original Monte Carlo swapping approach
• mip_mc - Monte Carlo selection with MIP optimization
• results - Result will be saved at here
• score - Grading function