Machine Learning with R
Download as PPTX, PDF8 likes2,850 views
The document presents a comprehensive introduction to machine learning using R, covering key topics such as exploratory data analysis, classification, clustering, and regression. It includes instructions for setting up R and RStudio, various machine learning algorithms, and techniques for performance evaluation and data visualizations. Additionally, it features practical demonstrations and links to resources for further exploration.
Machine Learning with R
- 1. Machine Learning with R Barbara Fusinska @BasiaFusinska
- 2. About me Programmer Machine Learning Data Solutions Architect @BasiaFusinska https://github.com/BasiaFusinska/MachineLearningWithR
- 3. Agenda ā¢ Whatās Machine Learning? ā¢ Exploratory Data Analysis ā¢ Classification ā¢ Clustering ā¢ Regression
- 4. Setup ā¢ Install R: https://www.r-project.org/ ā¢ Install RStudio: https://www.rstudio.com/ ā¢ GitHub repository: https://github.com/BasiaFusinska/Ma chineLearningWithR ā¢ Packages
- 7. Movies Genres Title # Kisses # Kicks Genre Taken 3 47 Action Love story 24 2 Romance P.S. I love you 17 3 Romance Rush hours 5 51 Action Bad boys 7 42 Action Question: What is the genre of Gone with the wind ?
- 8. Data-based classification Id Feature 1 Feature 2 Class 1. 3 47 A 2. 24 2 B 3. 17 3 B 4. 5 51 A 5. 7 42 A Question: What is the class of the entry with the following features: F1: 31, F2: 4 ?
- 9. Data Visualization 0 10 20 30 40 50 60 0 10 20 30 40 50 Rule 1: If on the left side of the line then Class = A Rule 2: If on the right side of the line then Class = B A B
- 10. Chick sexing
- 11. Supervised learning ā¢ Classification, regression ā¢ Label, target value ā¢ Training & Validation phases
- 12. Unsupervised learning ā¢ Clustering, feature selection ā¢ Finding structure of data ā¢ Statistical values describing the data
- 13. Supervised Machine Learning workflow Clean data Data split Machine Learning algorithm Trained model Score Preprocess data Training data Test data
- 14. Publishing the model Machine Learning Model Model Training Published Machine Learning Model Prediction Training data Publish model Test stream Scores
- 16. Classification problem Model training Data & Labels
- 17. Classification data Source #Links #Characters ... Fake TopNews 10 2750 ā¦ T Twitter 2 120 ā¦ F TopNews 235 502 ā¦ F Channel X 1530 3024 ā¦ T Twitter 24 70 ā¦ F StoryLeaks 722 1408 ā¦ T Facebook 98 230 ā¦ T ā¦ ā¦ ā¦ ā¦ ... Features Labels
- 18. Task: Iris EDA ā¢ Descriptive statistics (dimensions, rows, columns, data types, correlation) ā¢ Data visualization (distributions, outliers) ā¢ Features distributions & classes separation ā¢ 2D visualisation http://archive.ics.uci.edu/ml/datasets/Iris
- 19. K-Nearest Neighbours Algorithm ā¢ Object is classified by a majority vote ā¢ k ā algorithm parameter ā¢ Distance metrics: Euclidean (continuous variables), Hamming (text) ?
- 20. NaĆÆve Bayes classifier š š¶ š š) = š š¶ š š š š¶ š) š(š) š = (š„1, ā¦ , š„ š) š š¶ š š„1, ā¦ , š„ š) likelihood evidence prior posterior
- 21. NaĆÆve Bayes example Sex Height Weight Foot size Male 6 190 11 Male 6.2 170 10 Female 5 130 6 ā¦ ā¦ ā¦ ā¦ Sex Height Weight Foot size ? 5.9 140 8 š šššš š = š šššš š 5.9 šššš š 140 šššš š(8|šššš) šš£šššššš šš£šššššš = š šššš š 5.9 šššš š 140 šššš š 8 šššš + š šššššš š 5.9 šššššš š 140 šššššš š(8|šššššš) š šššššš š = š šššššš š 5.9 šššššš š 140 šššššš š(8|šššššš) šš£šššššš
- 22. Logistic regression š§ = š½0 + š½1 š„1 + āÆ + š½ š š„ š š¦ = 1 ššš š§ > 0 0 ššš š§ < 0 š¦ = 1 ššš š(š§) > 0.5 0 ššš š(š§) < 0.5 Logistic function Coefficients Best fit of Ī²
- 25. Evaluation methods for classification Confusion Matrix Reference Positive Negative Prediction Positive TP FP Negative FN TN Receiver Operating Characteristic curve Area under the curve (AUC) š“ššš¢šššš¦ = #ššššššš” #ššššššš”šššš = šš + šš šš + šš + š¹š + š¹š šššššš ššš = šš šš + š¹š š ššššš = šššš šš”šš£šš”š¦ = šš šš + š¹š šššššššššš”š¦ = šš šš + š¹š How good at avoiding false alarms How good it is at detecting positives
- 26. Task: Iris Classification ā¢ Data preprocessing ā¢ Split data for training and tests sets ā¢ Classification using: kNN and NaĆÆve Bayes ā¢ Performance evaluation ā¢ Results Visualisation
- 27. Task: Binary Classification ā¢ Only two slasses in the dataset (versicolor & virginica) ā¢ Classification using logistic regression ā¢ Performance evaluation ā¢ Results Visualisation
- 31. Data tuning Demo
- 32. Task: Resampling & Tuning ā¢ Repeated k-fold cross validation ā¢ Use NaĆÆve Bayes as classification algorithm ā¢ Tune the parameters using specific values ā¢ Performance evaluation
- 35. Hierarchical clustering ā¢ Decision of where the cluster should be split ā¢ Metric: distance between pairs of observation ā¢ Linkage criterion: dissimilarity of sets
- 36. Clustering Demo
- 37. Evaluating methods for clustering ā¢ Sum of squares ā¢ Class based measures ā¢ Underlying true
- 38. Task: Iris Clustering ā¢ Clustering using k-means and hierarchies ā¢ Compare clusters with the original classes assignments ā¢ Visualise the findings
- 39. Regression problem ā¢ Dependent value ā¢ Predicting the real value ā¢ Fitting the coefficients ā¢ Analytical solutions ā¢ Gradient descent š š = š½0 + š½1 š„1 + āÆ + š½ š š„ š
- 40. Ordinary linear regression Residual sum of squares (RSS) š š¤ = š=1 š (š¦š ā š„š š š¤)2 = š¦ ā šš¤ š š¦ ā šš¤ š¤ = ššš min š¤ š(š¤)
- 41. Task: Prestige EDA ā¢ Descriptive statistics (dimensions, rows, columns, data types, correlation) ā¢ Data visualization (distributions, outliers) ā¢ Handle missing data ā¢ Features significance
- 42. Evaluation methods for regression ā¢ Errors š šššø = š=1 š (šš ā š¦š)2 š š 2 = 1 ā (šš ā š¦š)2 ( š¦ ā š¦š)2 ā¢ Statistics (t, ANOVA)
- 43. Residuals vs Fitted ā¢ Check if residuals have non- linear patterns ā¢ Check if the model captures the non-linear relationship ā¢ Should show equally spread residuals around the horizontal line
- 44. Normal Q-Q ā¢ Shows if the residuals are normally distributed ā¢ Values should be lined on the straight dashed line ā¢ Check if residuals do not deviate severely
- 45. Scale-Location ā¢ Show if residuals are spread equally along the ranges of predictors ā¢ Test the assumption of equal variance (homoscedasticity) ā¢ Should show horizontal line with equally (randomly) spread points
- 46. Residuals vs Leverage ā¢ Helps to find influential cases ā¢ When outside of the Cookās distance the cases are influential ā¢ With no influential cases Cookās distance lines should be barely visible
- 48. Task: Prestige Regression ā¢ Numeric and categorical features ā¢ Other than linear relations ā¢ Combining the features
- 49. Categorical data for regression ā¢ Categories: A, B, C are coded as dummy variables ā¢ In general if the variable has k categories it will be decoded into k-1 dummy variables Category V1 V2 A 0 0 B 1 0 C 0 0 š š = š½0 + š½1 š„1 + āÆ + š½š š„š + š½š+1 š£1 + āÆ + š½š+šā1 š£ š
- 50. Categorical data for regression š š„ = š½0 + š½1 š„ + š½2 š£1 + āÆ + š½ š š£ šā1 + š½ š+1 š£1 š„ + āÆ + š½2šā1 š£ šā1 š„ š¦ ~ š„ + ššš” + š„: ššš”