SAMPATH-SEMINAR.pptx ..............................

1. The task of grouping data points based on their similarity with each other is called Clustering or Cluster Analysis. This method is defined under the branch of Unsupervised Learning, Clustering aims at forming groups of homogeneous data points from a heterogeneous dataset.

2. Discovering underlying patterns or groupings in data. Reducing the complexity of large datasets by summarizing them into groups. Market segmentation (e.g., grouping customers with similar buying patterns). Anomaly detection (finding outliers in data). Image compression (reducing color complexity by grouping similar pixel values).

3. k-Means is a popular partition-based clustering algorithm. Partitions data into ‘k’ clusters . Iteratively assigns points to clusters based on distance to centroids . Minimizes the sum of squared distances to the nearest centroid.

4. Step 1: Initialize Centroids The first step in k-Means clustering is to randomly initialize ‘k’ centroids . Centroids are the representative points that will be the center of each cluster. A smarter way to initialize centroids is to use k-Means++ , which selects initial centroids in a way that is more likely to lead to better results (reduces the chance of poor initial clustering). Example: For a 2D dataset like the one in the previous image (with height and width), the algorithm randomly selects k initial points as centroids.

5. Step 4: Repeat Until Convergence After recalculating the centroids, repeat the process by assigning each point to the new closest centroid . The algorithm continues to iterate: Reassigning points to the nearest centroid. Recalculating centroids for the new clusters. This process stops when convergence is reached, i.e., when centroids no longer move significantly between iterations or when the maximum number of iterations is reached. The algorithm guarantees convergence, though the final clustering depends on the initial random centroids.

6. Metrics help in:Quantifying the quality of clusters. Comparing different clustering algorithms. Choosing the right number of clusters (k). RMSE Silhouette Score Accuracy Precision, Recall, F1-Score ROC-AUC (if labeled data is available)

7. RMSE Definition : It measures the average squared distance between each data point and the centroid of its assigned cluster. Purpose : RMSE quantifies how well the data points fit within their clusters. Lower RMSE values indicate that points are closer to their centroids, meaning better clustering. Formula :

8. Definition : The Silhouette Score measures how similar a data point is to its own cluster compared to other clusters. Range : +1 : Data point is well matched to its own cluster. : Data point is on the boundary between clusters. -1 : Data point is likely misclassified into the wrong cluster. Formula :

9. Precision : Definition: Measures the accuracy of positive predictions. Formula: Precision = True Positives / (True Positives + False Positives) Importance: High precision indicates a low false positive rate. Recall : Definition: Measures the ability to find all relevant instances (true positives). Formula: Recall = True Positives / (True Positives + False Negatives) Importance: High recall indicates a low false negative rate. F1-Score : Definition: The harmonic mean of precision and recall, balancing both metrics. Formula: F1-Score = 2 * (Precision * Recall) / (Precision + Recall) Importance: Useful for imbalanced datasets where both false positives and false negatives matter.

10. Definition : The ROC curve visualizes a model’s performance by plotting the True Positive Rate (TPR) against the False Positive Rate (FPR) at various classification thresholds. Axes : X-axis : False Positive Rate (FPR) - the proportion of negative instances incorrectly classified as positive. Y-axis : True Positive Rate (TPR) - the proportion of positive instances correctly classified.

11. Definition : AUC provides a single scalar value that summarizes the model's ability to distinguish between classes across all thresholds. Interpretation : AUC = 0.5 : No discrimination ability (random performance). AUC > 0.5 : Some ability to differentiate between positive and negative classes. AUC = 1.0 : Perfect classification with no errors.

12. Problem Statement: Heart Disease Prediction In a healthcare setting, early detection of heart disease can significantly improve patient outcomes. A model is needed to predict whether a patient has heart disease based on several health indicators. Solution: Data Collection : Use the UCI Heart Disease dataset, which contains patient data such as age, cholesterol levels, blood pressure, and more. Data Preprocessing : Clean the dataset, handle missing values, and split it into training and test sets. Model Selection : Use a logistic regression model for binary classification (presence or absence of heart disease). Model Evaluation : Use the ROC-AUC metric to evaluate the model's performance.

13. What are some practical applications of k-means clustering in industry? How do you determine the best number of clusters (k) for a given dataset? What challenges might arise when using k-means clustering with large datasets? Can you explain how k-means clustering handles outliers?