Lab | Supervised Learning Model Evaluation

Introduction

You have already been familiar with the complete ML pipelines (both supervised and unsupervised) by conducting past labs. However, every dataset is different and as your experience grows you are able to choose better solutions in different scenarios. Therefore, keep practicing with all the datasets you can find as much as you can.

Linear regression model is not the silver bullet for all supervised learning analysis. In this lab we will present you a problem scenario where different supervised learning models are more appropriate. You will conduct a complete supervised learning analysis, apply different models, and compare their performances.

Getting Started

Open the main.ipynb file in the your-code directory. Follow the instructions and add your code and explanations as necessary. At the end, in addition to completing the cells please also save your RF model as a pickle file.

Deliverables

• main.ipynb with your responses.
• mushroom.sav file of your RF model.

Submission

Upon completion, add your deliverables to git. Then commit git and push your branch to the remote.

Resources

Mushroom Data Set @UCI MLP

Mushroom Classification @Kaggle

Consequences of multicollinearity

Chi-Square Test of Independence

pandas.crosstab

scipy.stats.chi2_contingency

pandas.get_dummies

sklearn.model_selection.train_test_split

Random Forest

Bagging and Random Forests

Support Vector Machine

sklearn.ensemble.RandomForestClassifier

Confusion Matrix

sklearn.metrics.confusion_matrix

Gradient Boosting

sklearn.ensemble.GradientBoostingClassifier

pickle - Python object serialization

Analysis and /classification of Mushrooms

The Search for Categorical Correlation