EBUS537 Data Mining and Machine Learning Assignment 1 | University of Liverpool
| Category | Assignment | Subject | Computer Science |
|---|---|---|---|
| University | University of Liverpool | Module Title | EBUS537 Data Mining and Machine Learning |
Assignment Requirements:
Questions
You are given a training dataset and a testing dataset, which will be provided in electronic form in Canvas (“myCarTrainDataset_2024.csv” and “myCarTestDataset_2024.csv”). Both datasets do not require data cleaning for simplicity. Each data object is a record of a car.
The attributes of the cars in the datasets are described below:
- “price”: purchasing price “doors”: number of doors
- “persons”: car capacity to accommodate persons
- “boot”: the size of luggage boot “accept”: car acceptability. This attribute is treated as the class label.
You are required to:
- Use the training dataset, conduct relevant data exploration to obtain initial insights into the dataset with appropriate explanations and interpretations.
- Use the training dataset, and apply the Hunt’s Algorithm together with the Greedy Strategy using the Gini impurity measure to build a fully-grown decision tree to predict whether a car is acceptable or not. If the attribute has multiple attribute values, you are required to use multiway split (do not use binary split). Leaf nodes should be declared as a single class label by applying the voting system as explained during the lectures (do not use probability/fraction). The selection of the attribute to split the decision tree should be explicitly explained and justified using the calculated results for the entire tree. The sample calculation processes and explanations should be provided as appropriate.
- After building the fully-grown decision tree in the previous step, please post-prune the sub-trees if all of its leaf nodes have the same class label if applicable. Test the post pruned decision tree using the test dataset and produce the confusion matrix. Interpret the obtained results in the case context.
- Beyond the above context, identify a case study of applying decision treebased classification methods in practice. Discuss the identified case study in relation to the CRISP-DM model. Support your arguments with relevant references
