j-que / ml-for-cvrp Goto Github PK

This project uses three different machine learning methods to train a model that can predict that best algorithm to use
out of a portfolio of algorithms on an instance of the capacitated vehicle routing prolem, based on 23 features of the
instance. The three machine learning methods used were random forest, k-nearest neighbors, and a single layer perceptron 
artifial neural network. The solving algorithms used were sef organizing map, genetic algorithm, the clarke and wright's
saving algorithm, and the sweep algorithm. The three models were created in the models folder. The data comes in from 
the cvrp_dataset csv file. The analysis folder contains jupyter notebooks were analyssi on the data and results of the 
models was done, along with a folder containg plots of the results.