k-Nearest Neighbor (kNN) classification
Tasks
1. Identify and describe the characteristics of the dataset (given) for you to work with for this
assignment (e.g., original resource information, dimension, etc.)
2. Execute the two kNN classification codes, including (1) [Link] (on Python command-line)
and (2) [Link] (on Google Colab).
3. Get an initial accuracy result without applying any sophisticated strategies.
4. Figure out what strategies you can apply to improve the performance.
5. Explain how you can specifically tailor the learning strategies for the data.
6. If possible, visualize your result in a couple of ways of visualizations, which you can find
from the references (given) as well as from other resources.
7. Write a short report that describes your findings.
Dataset
1. Pima Indians Diabetes data (“[Link]”)
2. The data file is given in the assignment directory so that you can run [Link] with the data
as given.
3. Where can you find the original data?
4. Briefly describe the characteristics of the original data. What is the shape? How many
classes? How many instances, features, and classes?
What should you do?
1. Briefly explain about the data.
2. Notice that the minimal python code ((1) [Link]) is given for you to start with. Therefore,
you can just use it as it is. However, you are welcome to extend it as you wish in order to do
your experimental study.
3. Based on (1) [Link], (2) [Link] can be generated. You need to figure out how to read
the data file from Google Colab. For example, you can read the data file directly from the
original data repository. Another way is to first store the data file into Google Drive and
read it using the file path in Google Drive.
4. Find out how you can improve the overall accuracy performance of kNN classification on
the data set. For example, you can investigate the classification accuracy performance with
varied parameter values, which include feature engineering, the number of nearest
neighbors (i.e., k), cross-validation split, etc.
5. Visualize your result if possible and briefly discuss the findings if you have any.
6. Record any heuristics or tips you learned from other resources and the experimental study.
Due date
Unless otherwise mentioned, every hands-on assignment is due one week after the posting.
Grading
The total point of this assignment is 100.
Report format and structure
The report length should be 3~5 pages (including references and possibly figures). When writing
the report, make sure to include the following:
1. Abstract (optional)
2. Introduction
3. Related work (optional)
4. Methodology (including (1) [Link] and (2) [Link] as well as other approaches you
applied)
5. Data (brief description of the given data file.)
6. Result
7. Discussion
8. Conclusion
9. Future work (if you have any, optional)
10. Reference (include the ones given below. Also, you are welcome to add more as you wish.)
Reference
1. [Link]
learn
2. [Link]
51209555453a
3. [Link]
4. [Link]
5. Step by Step Diabetes Classification-KNN-detailed
[Link]
6. A Complete ML Pipeline Tutorial (ACU ~ 86%)
[Link]