Description
Overview
In this project, you will implement Naive Bayes and K-Nearest Neighbour (K-NN) classifiers. You
will explore inner workings and evaluate behavior on a data set of stroke prediction, and on this basis
respond to some conceptual questions.
Implementation
The lectures and workshops contained several pointers for reading in data and implementing your
Naive Bayes classifier. You are required to implement your Naive Bayes classifier from scratch. You
may use Python libraries of your choice for implementing K-NN, evaluation metrics, procedures and
data processing.
For marking purposes, a minimal submission should have a preprocess() function, which opens
the data file, and converts it into a usable format. It should also define the following functions:
• split_data(), where you split your data sets into a training set and a hold-out test set.
• train(), where you build Naive Bayes and K-NN classifiers from the training data. You can
create train your data as you answer the related question.
• predict(), where you use a trained model to predict a class for the test data. You can also
do prediction as you answer the related question.
• evaluate(), where you will output the accuracy of your classifiers, or sufficient information
so that it can be easily calculated by hand.
There is a sample iPython notebook S2020S2-proj1.ipynb that summarizes these, which you
may use as a template.
You may alter the above prototypes to suit your needs; you may write other helper functions as you
require. Depending on which answers you choose to respond, you may need to implement additional
functions.
Packages
To provide the possibility of running your codes by markers, limit utilising the packages to the following packages in this project:
• pandas to read, split and preprocess the data
• sklearn to develop K-NN model and evaluate models
• numpy to implement scientific computing
• math to access to the mathematical functions
• matplotlib to create plots and visualizations
Data
For this project, we have adapted the Stroke data that have been used for stroke prediction [1], available online at (https://data.mendeley.com/datasets/x8ygrw87jw/1):
Some critical information:
1. File name: stroke_update.csv
2. 2740 instances
3. 10 attributes that include numeric and nominal attributes. The attributes avg_glucose_-
level, bmi and age are numeric. the rest of attributes are nominal.
4. The file stroke_features.txt explains each attribute
5. 2 classes, corresponding to the stroke outcomes: {0: No stroke, 1: Having stroke}
Questions
You should respond to questions 1-3. In question 2 (b) you can choose between two options for
smoothing and two options for Naive Bayes formulation. A response to a question should take about
100–200 words, and make reference to the data wherever possible.
Question 1
a Explore the data and summarise different aspects of the data. Can you see any interesting
characteristic in features, classes or categories? What is the main issue with the data?
Considering the issue, how would the Naive Bayes classifier work on this data? Discuss
your answer based on the Naive Bayes’ formulation [2]. (3 marks)
b Is accuracy an appropriate metric to evaluate the models created for this data? Justify
your answer. Explain which metric(s) would be more appropriate, and contrast their utility
against accuracy. [no programming required] (2 marks)
Question 2
a Explain the independence assumption underlying Naive Bayes. What are the advantages
and disadvantages of this assumption? Elaborate your answers using the features of the
provided data. [no programming required] (1 mark)
b Implement the Naive Bayes classifier. You need to decide how you are going to apply
Naive Bayes for nominal and numeric attributes. You can combine both Gaussian and
Categorical Naive Bayes (option 1) or just using Categorical Naive Bayes (option 2). Explain your decision.
For Categorical Naive Bayes, you can choose either epsilon or Laplace smoothing
for this calculation. Evaluate the classifier using accuracy and appropriate metric(s) on test
data. Explain your observations on how the classifiers have performed based on the metric(s). Discuss the performance of the classifiers in comparison with the Zero-R baseline.
(4 marks)
c Explain the difference between epsilon and Laplace smoothing. [no programming
required] (1 mark)
Question 3
a Implement the K-NN classifier, and find the optimal value for K. (1 mark)
b Based on the obtained value for K in question 4 (a), evaluate the classifier using accuracy
and chosen metric(s) on test data. Explain your observations on how the classifiers have
performed based on the metric(s). Discuss the performance of the classifiers in comparison with the Zero- R baseline. (2 marks)
c Compare the classifiers (Naive Bayes and K-NN) based on metrics’ results. Provide a
comparatory discussion on the results. [no programming required] (1 mark)
Submission
Submission will be made via the LMS, as a single Jupyter Notebook file. Submissions will open one
week before the submission deadline.
Assessment
5 of the marks will be assigned to whether the Python functions work correctly.
15 of the marks will be assigned to accurate and insightful responses to the questions, divided among
the questions as been stated in question descriptions. We will be looking for evidence that you have an
implementation that allows you to explore the problem, but also that you have thought deeply about
the data and the behavior of the relevant classifier(s).
Changes/Updates to the Project Specifications
If we require any (hopefully small-scale) changes or clarifications to the project specifications, they
will be posted on the LMS. Any addendums will supersede information included in this document.
Late submission penalty
There will be a penalty of 10% (2 marks) for every day (including working days and holidays) of late
submission.
Academic Misconduct
You are welcome — indeed encouraged — to collaborate with your peers in terms of the conceptualization and framing of the problem. For example, what the project is asking you to do, or what you
would need to implement to be able to respond to a question.
However, sharing materials beyond your group — for example, plagiarizing code or colluding in
writing responses to questions — will be considered cheating. We will invoke University’s Academic
Misconduct policy (http://academichonesty.unimelb.edu.au/policy.html) where
inappropriate levels of plagiarism or collusion are deemed to have taken place.
Data references
[1] T. Liu, W. Fan and C. Wu. A hybrid machine learning approach to cerebral stroke prediction based on imbalanced medical dataset. Artificial Intelligence in Medicine, volume.
101, No. 101723, 2019, DOI 10.1016/j.artmed.2019.101723
[2] M. A. Munoz, L. Villanova, D. Baatar and K. Smith-Miles. Instance spaces for machine learning classification. Machine Learning, volume. 107, pp. 109-147, 2018, DOI
10.1007/s10994-017-5629-5