Description
Given: Meal Data of 5 subjects
Amount of carbohydrates in each meal
Todo:
a) Extract features from Meal data
b) Cluster Meal data based on the amount of carbohydrates in each meal
First consider the given Meal data. Take the first 50 rows of the meal data. Each row is the meal amount of the corresponding row in the mealDataX.csv of every subject. So mealAmountData1.csv corresponds to the first subject. The first 50 rows of the mealAmountData1.csv corresponds to the first 50 rows of mealDataX.csv in Assignment 2.
Extracting Ground Truth: Consider meal amount to range from 0 to 100. Discretize the meal amount in bins of size 20. Consider each row in the mealDataX.csv and according to their meal amount label put them in the respective bins. There will be 6 bins starting from 0, >0 to 20, 21 to 40, 41 to 60, 61 to 80, 81 to 100.
Now ignore the mealAmountData. Without using the meal amount data use the features in your assignment 2 to cluster the mealDataX.csv into 6 clusters. Use DBSCAN or KMeans. Try these two.
Report your accuracy of clustering based on SSE and supervised cluster validity metrics.
Grading: I will give you a set of Meal data that is not included in the training set.
50 points for developing a code in Python or Matlab that takes the dataset and performs clustering
20 points for developing a code in Python or Matlab that implements a function to take a test input and run the clustering algorithm to provide the clustering result
30 points will be evaluated on the supervised cluster validation results obtained by your machine. This will be compared against class average to determine the final score.
Example:
O – 1 , 6, 9 10 1 – 3,4,9,11,12,15 >0 <= 20 >0 <= 20 3,4,5, 11, 12 13 2 – 1, 2, 10 0 >20 <= 40 2, 7,8, 14 15 3 – 5, 6, 7,8,14 >20 <=40
Classification error supervised cluster validity metric
2 + 1 + 2 = 5
Error 5/15 = 33.33 %
Test script that does KNN classification choose K, choose distance metric
Given a test data, calculate distances of the test data from each of your training data point.
Then do a K majority based classification
DBSCAN and K means
Test data input has 100 rows of mealDataX.csv
Output
Matrix
1 2
1 2
2 4
2 5
3 2
2 2
2 2
4 4
6 6
7 7
5 5
Output class numbering start from 1 and go till 6