Machine_Learning.pptx
- 3. Skill Set
- 4. Tool Set
- 7. ML in a Nutshell • Tens of thousands of machine learning algorithms • Hundreds new every year • Every machine learning algorithm has three components: • Representation • Evaluation • Optimization
- 8. Representation (Algorithms) • Decision trees • Sets of rules / Logic programs • Instances • Graphical models (Bayes/Markov nets) • Neural networks • Support vector machines • Model ensembles • Etc.
- 9. Evaluation(Model fit stats calculation) • Accuracy • Precision and recall • Squared error • Likelihood • Posterior probability • Cost / Utility • Margin • Entropy • K-L divergence • Etc.
- 10. Optimization(Methods for Parameters Estimation) • Combinatorial optimization • E.g.: Greedy search • Convex optimization • E.g.: Gradient descent • Constrained optimization • E.g.: Linear programming
- 11. ML in Practice • Understanding domain, prior knowledge, and goals • Data integration, selection, cleaning, pre- processing, etc. • Learning models • Interpreting results • Consolidating and deploying discovered knowledge • Loop
- 13. Machine Learning • How did our brain process the images? • How did the grouping happen? • Human brain processed the given images - learning • After learning the brain simply looked at the new image and compared with the groups classified the image to the closest group - Classification • If a machine has to perform the same operation we use Machine Learning • We write programs for learning and then classification, this is nothing but machine learning
- 15. The machine learning framework • Apply a prediction function to a feature representation of the image to get the desired output: Slide credit: L.
- 16. The Machine Learning framework y = f(x) output prediction function Image feature • Training: given a training set of labeled examples {(x1,y1), …, (xN,yN)}, estimate the prediction function f by minimizing the prediction error on the training set • Testing: apply f to a never before seen test example x and output the predicted value y = f(x) Methods includes Parametric Methods and Non-parametric Methods
- 17. Classification: predicting a category Some techniques: - Naïve Bayes - Decision Tree - Logistic Regression - SGD - Support Vector Machines - Neural Network - Ensembles
- 18. • Decision Trees – rpart, party, tree, chaid • Random Forest – randomForest, party • Neural Networks – nnet, neuralnet, RSNNS • SVM and Kernel estimation – e1071, kernalab •Performance Evaluation - ROCR Etc. Classification in R
- 19. Highest market share Data handling and analytics capabilities Easy to use/ learn Why R/Python
- 20. Table of Contents 1) Business Problem 2) Base knowledge 3) Difference between Linear Regression & Logistic Regression 4) Logistic Regression 5) Logistic Model Development 6) Model Validation 7) Hands on – R/Python 8) Quizes 9) Case Solution
- 22. Business Problem We have data for 900 banking customers, whether an applicant was rated as “Good Credit/Non-Default” or “Bad Credit/Default” based on some given attributes. Analytical Problem: We need to analyze and do exploratory data analysis and build model Here is the Definitions of the columns of the data: ⮚ Age-Age of Applicant ⮚ Education- Range 1-5 ⮚ Employment – Range 0-33 ⮚ Address – Range 0-34 ⮚ Income – Monthly Salary(‘Thousands) ⮚ Debtinc – Debt-to-Income Ratio ⮚ Credit debt – Credit card debt(credit loans) ⮚ Other debts – Any other loan liabilities ⮚ Default – 1(Loan default), 0 (Non-Default) 1. Understanding business problem 2. What is Default/Non-Default 3. Need of predictive model 4. Other problems that can be answered by similar scenario 5. What is Target/Dependent Variable 6. What are independent variables 7. Exploratory Analysis 8. Understanding Logistic Regression
- 24. Models can be developed in domains of risk, marketing or collections. • Risk: Models to assess the credit worthiness of an application ✓ Credit card & Loans: Bad vs. Good ✓ Fraud Detection: Fraud Vs. Not Fraud • Collections: Models to identify cardholders that are likely to default and thus need collection effort (Payment Projection Models) ✓ Credit card & Loans: Default vs. Non Default • Insurance: Models to identify claims that are Fraud or Not Fraud • Marketing & Sales: Models to identify to responders to promotional campaigns ✓ Marketing: Response Vs. Non Response ✓ Sales: Buying Vs. Not Buying • Operations: Models to identify to employees who attrite ✓ Operations: Attrition Vs. Retention • Website: Models to identify to weather visitor will click or not ✓ Website: Click Vs. Not Click • Gaming: Models to identify to who will win ✓ Gaming: Win Vs. Loss • Health Care: Models to identify to cure or not cure ✓ Health Care: Cure Vs. Not Cure Logistic Regression Modeling Applications
- 25. Meaning of Modeling By “Modeling” we mean developing set of equations or mathematical formulation by which we can : • Predict certain events • Identify the drivers of certain events based on some explanatory variables For example, we can build models to predict drivers of sales, risk of a borrower. Historical Data Statistical Analyse s Predict Future Events Bad 25 Your Company
- 26. Recall: Modeling Process 3) Model Development • Multivariate Analysis • Variable Creation • Variable Transformation and Binning • Multicollinearity Check • Variable Selection • Technique Selection Building Blocks of the Modeling Process 2) Data Analysis • Univariate Analysis • Extreme value treatment • Missing Value Treatment • Graphical Representation 1) Data Preparation • Data Extraction • Data Collation • Data Sampling and holdouts 4) Model Validation • Performance Analysis and charting • Multiple model comparison • Out of Sample Validation
- 27. What ? Regression is formulating a functional relationship between a set of independent or Explanatory variables (Xs) with a Dependent or Response variable (Y). Y=f(X1, X2, X3,…,Xn) Why ? ▪Knowledge of Y is crucial for decision making. ▪ Will he/she buy or not? ▪ Shall I offer him/her the loan or not? ▪X is available at the time of decision making and is related to Y, thus making it possible to have a prediction of Y. Recall: What is Regression?
- 28. ÐLogistic Regression Continuous Binary (1/0) ÐOrdinary Least Squares Regression Y Sales Volume, Claim Amount, Number of Equipments, % Increase in Growth, etc. Buy/No Buy, Delinquent/Non Delinquent, Growth/No Growth, etc. Recall: Types of Regression
- 29. Quiz 1 Is Logistic regression mainly used for Regression? True False A B 1. A 2. B
- 31. Why Not Linear?
- 34. Quiz 2 Which of the following methods do we use to best fit the data in Logistic Regression? Least Square Error Maximum Likelihood A B Jaccard distance Both A and B C D 1. A 2. B 3. C 4. D
- 36. Steps of Modelling Outlier treatment •Box Plot •Percentile(99) Missing value treatment Normalizatio n of data Handling multi- collinearity •Performed VIF Dummy variable creation •Handle categorical variables Logisitc Model • Variable Selection • Optimising Model
- 37. EDA on banking customer data
- 38. Missing Value Analysis Console Output
- 39. Outlier Detection: identify abnormal patterns Example: identify engine anomalies Features: - Heat generated - Vibration of engine
- 40. Outlier Detection Target Function: outlier factor Outlier factor (0…1) Some techniques: - Statistical techniques - Local outlier factor - One-class SVM ID Total$ Age City OF 101 $200 25 SF 0.1 102 $350 35 LA 0.05 103 $25 15 LA 0.2 … … … … 0.1 0.9 0.2 0.15 0.1
- 41. Quiz 3 Which of the following measures of central tendency will always change if a single value in the data changes? Mean Median A B Mode All C D 1. A 2. B 3. C 4. D
- 43. Quiz 4 A correlation between age and health of a person found to be -1.09. On basis of this you would tell the doctors that? The age is good predictor of health The age is poor predictor of health A B None of these C 1. A 2. B 3. C
- 47. Logistic Procedure-Concordance Association of Predicted Probabilities and Observed Responses Percent Concordant 79.0 Percent Discordant 19.1 Percent Tied 1.9 Pairs 3627468 ✓ Concordance is used to assess how well scorecards are separating the good and bad accounts in the development sample. ✓ The higher is the concordance, the larger is the separation of scores between good and bad accounts. ✓ The concordance ratio is a non-negative number, which theoretically may lie between 0 and 1. Concordance Determination: Among all pairs formed from 0 & 1 observations from the dependent variable, the % of pairs where the probability assigned to an observation with value 1 for the dependent variable is greater than that assigned to an observation with value 0. Percentage of concordant pairs should be at least greater than 60.
- 48. Logistic Procedure-Confusion Matrix/ROC Curve
- 49. Solution 1. AUC of the model being 87% means that model have been able to distinguish between defaulters and non-defaulters with 37% better than baseline/random model 1. Initial event rate i.e. default rate being 26% which is baseline result while Precision of model being 55%, model have been able to provide 2x better lift.
- 50. Quiz 5 For the below confusion matrix, what is the recall? 1. 0.7 2. 0.8 3. 0.9 4. 0.95
- 51. Q & A
- 53. Thank You