MLProject1 - Higgs Boson

Machine Learning Project 1

Team Members

Serif Soner Serbest: serif.serbest@epfl.ch

Aim :

In this project we predict CERNs simulated particle collision events as either a Higgs Boson signal or background noise as of binary classification, which is a Kaggle

Result:

We ranked 28th in Kaggle LeaderBoard among 211 teams,with our score: 0.83224.

Run

To get the exact results run the "run.py" file.

Functions

Loading data

labels,features,data,id_ = load_csv_data(data_path, sub_sample=False)

Divide training data into 3 subsets

3 subsest according to Jet category = categorize_data(prediction, data)

Process data and build model

Cleaned and standartized features = process_data(features) build_model_data(prediction, x)

Apply Polynomial Expansion

Polynomially Expanded Features = build_poly(features, degree) :

Apply Regularized Logistic Regression

weight, loss = reg_logistic_regression(labels,feature_model,lambda,max iteration, gamma)

Apply Ridge Regression

weight, loss = ridge_regression(labels, feature_model, lambda)

Label predictions

prediction = predict_labels(weight,feature_model)

Merge all three categories

labels = decategorize_prediction(row_size, label1, label2, label3, indices1, indices2, indices3)

Create submission file

create_csv_submission(test_id_,labels,"submission.csv")