A multi-parameter, multi-model k-fold grid search. Designed to find the most promissing model for your classification problem.

A template to import, treat and export data in the correct format. Highly dependent on your goals.

The classification module containing the classification class.

Current compatible classifiers

• SVC
• DecisionTreeClassifier
• KNeighborsClassifier
• LogisticRegression
• GaussianNB
• RandomForestClassifier
• Perceptron
• SGDClassifier
• XGBClassifier

A script which joins data_preprocessing and GSClassification modules, in order to return statistics about each of the evaluated classifiers and their best parameters.

You need to set:

• Classifiers (comment out the ones you do not wish to include)
• Grid search parameters

This software comes with an example from Kaggle's introductory competition. It will preprocess data, test different parameters for different models (executing k-fold grid searches to avoid overfitting) and return

• a sorted pandas dataframe with the best accuracies and parameters;
• the cumulative accuracy profile (CAP) along with an image representation of the above metrics.

In order to modify parameters of the grid search, check sklearn's GridSearchCV guide.