Identifying Legendary Pokemon using Machine Learning

A simple project documenting basic steps to build a simple machine learning model using the most loved pokemon as an example! Our task is to identify which pokemon are the LEGENDARY ones. With some explanation of different classifying algorithms to choose from! Sounds cool???

Transcript

1. Identifying Legendary Pokemon using Machine Learning Yip Mei Yee @ DHL IT Services
2. What is machine learning?
3. Application Fraud Detection Stock Price Prediction Identifying Disease and Diagnosis
4. Data Science Process Data Collection Data Exploration Data Cleaning Data Transformation Modelling Evaluation
5. Legendary Pokemon?
6. Data Collection
7. Data Import
8. First Step, check out the data.
9. Name of Each Pokemon Each pokemon has a type, this determines weakness/resistance to attacks Some pokemon are dual type and have 2 a general guide to how strong a pokemon is hit points, or health the base modifier for normal attacks special attack damage resistance against special attacks determines which pokemon attacks first each round base damage resistance against normal attacks Number of Generation Is this Pokemon Legendary or not
10. Feature & Label Features Target Label
11. Data Cleaning Why is it so important?
13. Rename column
14. Drop Duplicates
15. Handling of Null Values
16. Drop Unnecessary Column
17. Data Analysis/ Exploration
19. Data Transformation
20. One Hot Encoding
21. Data Normalization
22. One Hot Encoding of Categorical Data
23. Data Train (80%) Test(20%) Data Splitting
24. Data Splitting
25. Modelling
26. Logistic Regression
27. Logistic Regression
28. Decision Tree
29. Decision Tree
30. Random Forest
31. Random Forest
33. Naïve Bayes
34. K-Nearest Neighbour
35. K-Nearest Neighbour
37. Training Done. What’s next?
38. Evaluate Model Performance
40. Accuracy Overall, how often is our model correct?
41. Precision When the model predicts positive, how often is it correct?
42. Recall when it’s actually yes, how often does it predict yes.
43. AUC-ROC Curve ROC: Probability Curve AUC: Degree of separability It tells how much model is capable of distinguishing between classes.
44. Test Results
45. Confusion Matrix of Decision Tree Non Legendary Pokemon Legendary Pokemon Non Legendary Pokemon 136 1 Legendary Pokemon 1 7 Predicted Actual
46. Decision Tree
47. Conclusion 1) Decision Tree is the best classifier. 2) Building a model requires steps below: Data Collection Data Exploration Data Cleaning Data Transformation Modelling Evaluation
48. Questions?