rpotter6298
32 lines
2.4 KiB
Markdown
Executable File
32 lines
2.4 KiB
Markdown
Executable File
This assignment consists of a small project. The task is to classify data. What you need is a dataset that is suitable for classification. You can use your own dataset (from your studies, interests, hobbies, etc.) or a publicly available dataset, e.g. form scikit-learn.org, uci Links to an external site. or Kaggle Links to an external site..
|
|
In the first part of the assignment you will explore and describe the dataset and build a decision tree classifier and analyze its performance. In the second part of the assignment you will build an MLP classifier for the same task, analyze its performance and compare the performances and other pros and cons of the decision tree classifier and the MLP classifier.
|
|
Part 1: Exploratory Data Analysis and Decision Tree Classifier
|
|
|
|
Find the right dataset for your task
|
|
Identify and motivate your objective
|
|
Do an exploratory data analysis with jupyter notebook and report your findings.
|
|
Motivate why the data in the dataset is the right data for your task (classification task)
|
|
Iteratively find the best decision tree classifier for the task.
|
|
Steps to consider include but are not limited to
|
|
Restricting the depth of the tree using different options to ovoid overfitting and underfitting
|
|
Preprocessing the data in different ways
|
|
|
|
Always motivate what you do, why and how.
|
|
|
|
Present your best decision tree model and motivate why you think this is the best model for your task and show and describe how you evaluated this.
|
|
Your report in Jupyter notebook should report your step by step investigation of the problem and the iterative development of the decision tree classifier and its evaluation.
|
|
|
|
Part 2: MPL Classifier and comparison of Decision Tree and MLP classifier
|
|
|
|
Iteratively find the best MLP classifier for the task
|
|
Steps to consider include, but are not limited to
|
|
Data preprocessing
|
|
MLP hyperparameter tuning
|
|
|
|
Compare the performance of the Decision Tree Classifier from part 1 with the performance of the MLP classifier from part 2.
|
|
Investigate
|
|
The performance of the model
|
|
The difficulty of building the model
|
|
The transparency of the model
|
|
|
|
You need to hand in your notebook as an HTML. To create HTML file, you go to File/Save and export notebook as… and choose HTML. Check the created HTML before you upload it on CANVAS |