Skip to content

Latest commit

 

History

History
57 lines (40 loc) · 4.28 KB

README.md

File metadata and controls

57 lines (40 loc) · 4.28 KB
Raw

Machine Learning A-Z by SuperDataScience Team

This repository contains all the codes and datasets for Machine Learning A-Z Course, by Kirill Eremenko and Hadelin de Ponteves.

Verdict for using R & Python for the tasks

Objective R Python Verdict
Data Preprocessing Requires fewer packages Easy to impute missing values and splitting into training and test sets, has LabelEncoder and OneHotEncoder Python
Linear Regression Can be done in a single line Requires multiple lines to code R
Polynomial Regression Unless you use the poly function the process is cumbersome Easier to define degrees with PolynomialFeatures from sklearn.preprocessing Python
Support Vector Regression e1071 makes it very easy Requires feature scaling, therefore lengthy R
Decision Tree Regression Process requires to define control variable Easy going process with DecisionTreeRegresson from scikitlearn Python
Random Forest Regression Process requires to define control variable Same code as Decision Tree, easy process with sklearn.ensemble - RandomForestRegressor Python
Logistic Regression Requires multiple lines of code Much more intuitive Python
k-NN Classification Careful while plotting training and test sets Intuitive Python
SVM Ease of use Same ease of use R
Kernel-SVM Ease of use Same ease of use R
Naive-Bayes Classifier Careful during the call of naiveBayes function for x & y values Ease of use Python
Decision Trees Helpful with the plot and text function for classifier object Intuitive R
Random Forest Classifier Careful during the call of randomForest function for x & y values Ease of use with RandomForestClassifier Python
Clustering k-means is built-in in R and Hierarchical Clustering is easier Better for visualization R
Association Rule Learning Extremely easy and intuitive with built in functions for both Apriori & Eclat using the arules library Apriori is still intuitive but Eclat is a lenghty algorithm and complicated R
Reinforcement Learning using UCB & Thompson Sampling Complex algorithm and lenghty Relatively easier to implement Python
Natural Language Processing Extremely easy and short procedure with tm Fairly intuitive R
Artificial Neural Networks Fast process as it makes use of all the cores available. Computational difference through external server is phenomenal through h2o Takes time to process hidden layers R
Convlutional Neural Networks DeepWater thorugh h2o is in beta stage Step by step process Python
Dimensionality Reduction Long process, caret library is tricky to operate with In-built libraries make the work far more simpler Python
Model Selection and Boosting Code is complicated and reqiures multiple functions Relatively easier with built in functions from cross_val_score Python

Course Structure

  1. Data Preprocessing and Classification

  2. Simple Linear Regression

  3. Multiple Linear Regression

  4. Polynomial Regression (Contains Random Forest and SVR)

  5. Classification (Contains Logistic Regression, Decision Trees, k-NN, Naive Bayes, Random Forest, SVC, SVM, Kernel-SVM)

  6. Clustering (k-means and Hierarchical Clustering along with Elbow Representations)

  7. Association Rule Learning (using Apriori and Elclat Algorithms)

  8. Reinforcement Learning (using UCB and Thompson Sampling)

  9. Natural Language Processing (using nltk in Python and tm in R)

  10. Neural Networks (through ANN and CNN - Note, CNN through DeepWater is in beta stage and has not been implemented in R)

  11. Dimensionality Reduction (using Principal Component Analysis, Linear Discriminant Analysis & Kernel PCA)

  12. Model Selection and Boosting

Annex. 1 Containg various templates for Regression, Plotting, Classification, Cross Validation and Plotting along with PDFs from the lecture