This proyect was done for a university research workshop (TID) where the objetive is to compare the performance of different optimization solvers. The solvers compared are Gurobi in python and constrOptim in R.
The data used comes from the library Bayesm in R (OrangeJuice). It contains the sales data from 83 stores from the Chicago area for 11 refrigerated orange juice for 121 weeks. The optimization model uses store number 2, wanting to maximize the Category Revenue of orange juice deciding the price of each product. Machine learning is use to predict the demand considering the log of prices, log of demand, deal and feat status to traing the regression.
The file "Gurobi Explanation" explains how to use Gurobi given the optimzation model for OrangeJuice, going step by step to undestand Gurobi's functions.
The file "Comparison_solvers" is where the Gurobi optimization is done and compared with the constrOptim results which is in "ConstrOptim11products". "Comparison_solvers" also has the change from lineal regression to Lasso regression to predict the demand to get more realistic results after optimizing.
The CSV files "Pricesdf11OrangeJuice" and "demandadf11OrangeJuice" are files exported from python to have the same data to optimize with constrOptim in R.
"data11dfGraph_iter50" is a CSV file exported from R, this files has the incumbents found after optimizing the model with constrOptim running 50 iterations, it also has the time it took each iteration. This way we are able to compare the performance of the solvers.