ML basic & clustering

ML basic & clustering

• 1.
  What is DataScience • Also Known as data driven Science • It is an interdisciplinary field about scientific methods, processes and system to extract knowledge or insights from data in various form either structured or unstructured
• 2.
  What is MachineLearning  Machine Learning is a concept which allows the machine to learn from examples and experience, and that too without being explicitly programmed. So instead of you writing the code, what you do is you feed data to the generic algorithm, and the algorithm/ machine builds the logic based on the given data
• 3.
  Features of MachineLearning  It uses the data to detect patterns in a dataset and adjust program actions accordingly  It focuses the development of computer programs that can teach themselves to grow and change when exposed to new data  It enables the computer ti find hidden insights using interactive algorithms without being explicitly programmed  It is a method of data analysis that automate analytical model building
• 4.
  How ML ModelWorks
• 5.
  Application
• 6.
  Stages of MachineLearning
• 7.
  Machine Learning Algorithms Supervised LearningUnsupervised Learning ClusteringDecission Tree Random Forest KNN SVM Naïve BayesLogistic Regression Hierchical K-Means DB-Scan Linear Regression PCA
• 8.
  Clustering  It refersto the grouping of records, observations into classes of similar objects.  It is a collection of records that are similar to one another and dissimilar to records in other clusters.  No target variable for clustering  It segments the entire dataset into homogeneous subgroups where similarity within cluster is maximized and outside cluster is minimized
• 9.
  Hierchical Clustering Divisive MethodAgglomerative
• 10.
  Linkage Function There areseveral linkage functions available for hierarchical clustering We will focus on these three commonly used methods  Single linkage  Nearest Neighbour approach  Based on minimum distance between any records in two clusters  Complete linkage  Farthest Neighbour approach  Based on maximum distance between any records in two clusters  Average linkage Average distance of all the records in cluster.
• 11.
  Example of HierchicalClustering Consider A, B, C, D, E as cases with the following similarities: A B C D E A - 2 7 9 4 B 2 - 9 11 18 C 7 9 - 4 8 D 9 6 4 - 2 E 4 18 8 2 -
• 12.
  Example Contd.. So letscluster E and B. We now have the structure:
• 13.
  Example Contd..  Nowwe update the case-to-case matrix A BE C D A - 4 7 9 BE 4 - 9 11 C 7 9 - 4 D 9 6 4 - Note: To compute BE -- SC (A, B) = 2 SC (A, E) = 4 SC(A,BE) = 4 if we are using single linkage SC(A,BE) = 2 if we are using complete linkage SC(A,BE) = 3 if we are using group average
• 14.
  So lets clusterBE and C. We now have the structure:
• 15.
  Now we updatethe case-to-case matrix. A BCE D A - 7 9 BCE 7 - 2 D 9 6 - To compute SC(A, BCE): SC (A, BE) = 2 SC (A, C) = 7 so SC(A,BCE) = 2 To Compute: SC(D,BCE) SC(D, BE) = 2 SC(D, C) = 4 so SC(D, BCE) = 2 SC(D,A) = 9 which is greater than SC(A,BCE) or SC(D,BCE) so we now cluster A and D.
• 16.
   So letscluster A and D. At this point, there are only two nodes that have not been clustered, AD and BCE. We now cluster them.
• 17.
  Now we haveclustered everything
• 18.
  Advantage & Disadvantage Advantagesof hierarchical clustering are Easy to understand Often efficient in clustering Disadvantages of hierarchical clustering are  Not very much scalable  Choice of distance measure is far from trivial job  Not applicable for dataset with missing values  For huge dataset it wont work  Due to heuristic nature, greedy search may result in unclear cluster hierarchy
• 19.
  K-Means Clustering  Itis an algorithm to group the objects based on attributes/features into K number of group  Steps for K-Means Step 1: Begin with a decision on the value of k = number of clusters Step 2: Put any initial partition that classifies the data into k clusters. You may assign the training samples randomly,or systematically as the following • Take the first k training sample as single- element clusters • Assign each of the remaining (N-k) training sample to the cluster with the nearest centroid. After each assignment, recompute the centroid of the gaining cluster
• 20.
  Step 3: Takeeach sample in sequence and compute its distance from the centroid of each of the clusters. If a sample is not currently in the cluster with the closest centroid, switch this sample to that cluster and update the centroid of the cluster gaining the new sample and the cluster losing the sample Step 4 . Repeat step 3 until convergence is achieved, that is until a pass through the training sample causes no new assignments
• 21.
  Example of K-MeansClustering  Let us consider a simple dataset DataPoints Axis Points a (1,3) b (3,3) c (4,3) d (5,3) e (1,2) f (4,2) g (1,1) h (2,1)
• 22.
   Step 1:Randomly assign any two individuals as two centroids DataPoints Axis Points a (1,3) b (3,3) c (4,3) d (5,3) e (1,2) f (4,2) g (1,1) h (2,1)
• 23.
   Finding NearestCluster centre for each record Point Distance from m1 Distance from m2 Cluster Membership a 2 2.24 c1 b 2.83 2.24 c2 c 3.61 2.83 c2 d 4.47 3.61 c2 e 1.00 1.41 c1 f 3.16 2.24 c2 g 0 1.00 c1 h 1.00 0 c2
• 24.
   Finding NearestCluster centre for each record(second pass) Point Distance from m1 Distance from m2 Cluster Membership a 1.00 2.67 c1 b 2.24 0.85 c2 c 3.16 0.72 c2 d 4.12 1.52 c2 e 0.00 2.63 c1 f 3.00 0.57 c2 g 1.00 2.95 c1 h 1.41 2.13 c1
• 25.
   Finding NearestCluster centre for each record(third pass) As no records have shifted cluster membership, cluster centroids also remain unchanged & Algorithm terminates Point Distance from m1 Distance from m2 Cluster Membership a 1.27 3.01 c1 b 2.15 1.03 c2 c 3.02 0.25 c2 d 3.95 1.03 c2 e 0.35 3.09 c1 f 2.76 0.75 c2 g 0.79 3.47 c1 h 1.06 2.66 c1
• 26.
  Advantage & Disadvantage Advantages •Very simple algorithm • Always converges (even though locally) • Quite fast and interpretation of clusters is quite easy Issues • Greatly affected by extreme values • Performs poorly for irregular shaped data points (longitude and latitude) • Each time it is run, different results may come out! • Cannot handle categorical data
• 27.
  How to findCluster Goodness  Silhouette Score  For each of data value i, ai=distance between data value and its cluster centre bi=distance between data value and next closest cluster centre Cluster Characterstics: If Si>0.5, Good evidence of reality of clusters If Si<0.25 Bad evidence of reality of clusters If 0.25<Si<0.5, Some evidence of reality of the clusters