Skip to content

Latest commit

 

History

History

DataMining

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
Solution.ipynb
Solution.ipynb
 
 
SyntheticQ1.csv
SyntheticQ1.csv
 
 
seeds.csv
seeds.csv
 
 
stones.csv
stones.csv
 
 

README.md

Data Mining

Problem 1

In this problem, you are required to apply various clustering techniques on a given dataset SyntheticQ1.csv, which is an artificial dataset containing 4 convex clusters. The dataset has two attributes (X and Y) for each instance, delimited by semicolons.

  1. Data Preprocessing

    • Remove records with missing values (empty or marked with ‘?’).
    • Remove records where X or Y has a negative value.

  2. Clustering and Visualization

    • K-means Clustering: Apply K-means to generate 4 clusters.
    • Scatter Plot Visualization of the clusters.
    • DBSCAN Clustering: Use ε = 0.5 and minPts = 3 on the dataset.
    • Visualize DBSCAN clusters using a scatter plot.
    • Single-Linkage Hierarchical Clustering: Generate 4 partitions and visualize with a scatter plot.
    • Complete-Linkage Hierarchical Clustering: Generate 4 partitions and visualize with a scatter plot.
    • Average-Linkage Hierarchical Clustering: Generate 4 partitions and visualize with a scatter plot.

  3. Comparison

    • Briefly compare and explain the outcomes of all clustering techniques.

Note: Do not use dimensionality reduction during visualization.

Problem 2

Apply clustering techniques on the seeds.csv dataset, containing 4 attributes related to plant seeds: length, width, asymmetry coefficient, and compactness coefficient.

  1. Elbow Method: Determine the optimal number of clusters.
  2. K-means Clustering: Use the optimal number of clusters.
  3. Visualization and Heatmap
    • Scatter Plot Visualization with dimensionality reduction.
    • Heatmap of the clustering results.
  4. Hierarchical Clustering: Apply hierarchical clustering with K partitions (found from the elbow method).
  5. Comparison: Briefly compare K-means and hierarchical clustering outcomes.

Problem 3

Analyze stones.csv dataset, which includes height, width, density, compactness, and texture, with class labels (A to F).

  1. Classification Modeling

    • Split dataset into 60% train and 40% test.
    • Decision Tree: Build a model and generate a confusion matrix and classification report.
    • KNN (K=5): Build a model and generate a confusion matrix and classification report.
    • SVM (Polynomial kernel, degree 3): Build a model and generate a confusion matrix and classification report.

  2. New Entry Classification

    • Classify a new entry with specified feature values using all models.

Problem 4

Using the spambase.data dataset from UCI repository, perform 500 random splits (80% train, 20% test) and apply the following classification techniques:

  • Decision Trees
  • KNN
  • Support Vector Machines
  • Logistic Regression
  • Naïve Bayes

Generate a summary table showing average precision, recall, F1 score, and accuracy across the 500 splits.

Problem 5

Select a benchmark dataset, partition it into train and test sets, and apply classification techniques. For each classification technique, plot the variations in accuracy, precision, recall, and F1 score for different train/test ratios.

Dataset Source: UCI Repository