BRUG meetup (Bangalore R User Group)

This repository captures content of my presentations, code notebooks / information shared in Bangalore R User Group meetup events.

Bagging and Boosting in R

• Please refer to the deck Bagging_Boosting_in_R.pdf for presentation content.
• The related source file depicting comparision between algorithms using a sample dataset is demonstrated here

Interpretable Machine Learning / ML Explainability / XAI

• Please refer to the deck ML_Interpretability.pdf for presentation content.
• The related Jupyter notebook depicting sample dataset demonstration of a LIME package usage is captured here The nbviewer version is here which will display the charts for some LIME explanations appropriately.
• Following categorization can be looked at from Model Explainability and Interpretability standpoint:
  • Transparent models (Linear / Logistic Regression, Decision trees, k-Nearest Neighbours, Rule based learners, Bayesian models, Generative additive models)
  • Opaque models (Random forests, SVM, Multi layer Neural Net)
  • Both of these Transparent and Opaque models have Model agnostic and Model specific explainability categories
    • Model Agnostic
      • Explanation by simplification
        • Rule based learner
        • Decision tree
      • Feature relevance explanation
        • Influence functions
        • Sensitivity
        • Game theory inspired (SHAP)
        • Interaction based
      • Local explanations
        • Rule based learner (Anchors)
        • Linear approximation (LIME)
        • Counterfactuals
      • Visual explanations
        • Sensitivity
        • Dependency plots (ICE, PDP etc.)
    • Model Specific
      • Explanation by simplification
        • Rule based learners (InTrees)
        • Decision trees (InTrees)
        • Distilation
      • Feature relevance explanation
        • Feature importance

10+10 use cases in Retail in 2020

• Please refer to 10plus10_in_Retail_in_2020.pdf for the presentation.
• The purpose is to present predictions around key business use cases that will continue to dominate in Retail in Data Science and AI in 2020. Stakeholders should focus these use cases to get benefits and create impact and value for their business.

Extreme Gradient Boosting in R

• The purpose is to understand concepts of the scalable tree boosting approach (XGBoost) in R, it's features etc. This solves many data science problems in relatively fast and accurate manner.
• The code snippet is under here
• Please refer to XGBoost_in_R.pdf for the presentation.
• DIsclaimer: Some of this content, approach are reused from various references.

Deep Learning using R

• Please refer to the DeepLearning_using_R.pdf for details.
• Objective is to provide a very high level view about Deep learning and some packages in R. Example of MNIST dataset can be leveraged to showcase the use of R using various libraries.

Packages / Libraries in R

• Please refer to Packages_and_OOP_in_R.pdf deck for content on the presentation conducted as part of BRUG. This focuses on Packages and OOP in R (at a high level).

• The following list does not involve entire exhaustive list. However, the intent is to provide some key and important packages that are used and helpful in most CRISP-DM phases.

R Packages at a glance by category

Package category	Package Name	Features
Data Manipulation	dplyr	Data wrangling, working with remote data frames
Data Manipulation	data.table	Data aggregation involving large datasets, file reader and parallel file writer
Data Manipulation	lubridate	Working with date and time formats, parsing of date-time data
Data Manipulation	jsonlite	Robust parsing of JSON objects in R

Package category	Package Name	Features
Graphic Display	ggplot2	Powerful implementation of the grammar of graphics visualization, Plot specifications
Graphic Display	corrplot	Abilities to visualize correlation matrices and confidence intervals
Graphic Display	lattice	Emphasis on multivariate data

Package category	Package Name	Features
HTML Widget	plotly	Rich features around charts, web based toolbox for building visualizations
HTML Widget	ggvis	Implementation of an interactive grammar of graphic
HTML Widget	DT(DataTables)	Displays R matrices and data frames as interactive HTML tables
HTML Widget	rCharts	Interactive JS charts from R

Package category	Package Name	Features
Reproducible Research	knitr	Easy dynamic report generation in R, enables integration of R code into LaTex, HTML, Markdown, AsciiDoc, reStructuredText documents
Reproducible Research	rMarkdown	Next generation implementation of R Markdown based on pandoc
Reproducible Research	slidify	Generated reproducible html5 slides from R markdown

Package category	Package Name	Features
Machine Learning	mlr	Extensible framework for classification, regression, survival analysis and clustering, easy extension mechanism through S3 inheritance
Machine Learning	xgboost	Implementation of Gradient Boosted Decision Trees algorithm
Machine Learning	caret	Multiple model comparision and usage for classification and regression
Machine Learning	gbm	Generalized Boosted Regression Models
Machine Learning	prophet	Forecast for time series data, manages data with multiple seasonality with linear or non-linear growth
Machine Learning	randomforest	Implements Breiman's random forest algorithm for classification
Machine Learning	Arules	Mining Association Rules and Frequent itemsets
Machine Learning	Boruta	Wrapper algorithm for all relevant feature selection
Machine Learning	Forecast	Timeseries forecasting using ARIMA, ETS, STLM, TBATS, and neural network models
Machine Learning	Anomalize	Tidy Anomaly Detection using Twitter’s AnomalyDetection method
Machine Learning	AnomalyDetection	AnomalyDetection R package from Twitter
Machine Learning	e1071	Misc Functions of the Department of Statistics (e1071)
Machine Learning	MXNet	MXNet brings flexible and efficient GPU computing and state-of-art deep learning to R

Package category	Package Name	Features
Web Search	Rcurl	general network (HTTP/FTP…) client interface for R
Web Search	Curl	flexible web client for R
Web Search	Httr	user friendly Rcurl wrapper
Web Search	shiny	simple interactive web applications with R
Web Search	Plumber	A library to expose existing R code as web API
Web Search	Rfacebook	access to facebook API via R

Package category	Package Name	Features
Database Management	RODBC	ODBC database access for R
Database Management	DBI	common interface between R and DBMS
Database Management	Elastic	wrapper for elastic search HTTP API
Database Management	ROracle	OCI based Oracle database interface for R
Database Management	RPostgreSQL	R interface to PostgreSQL database system
Database Management	RSQLite	SQLite interface for R

Package category	Package Name	Features
NLP Specific	text2vec	Fast Text Mining Framework for Vectorization and Word Embeddings
NLP Specific	tm	A comprehensive text mining framework for R
NLP Specific	OpenNLP	Apache OpenNLP Tools Interface
NLP Specific	koRpus	An R Package for Text Analysis
NLP Specific	LDAvis	Interactive visualization of topic models
NLP Specific	SnowballC	Snowball stemmers based on the C libstemmer UTF-8 library
NLP Specific	Tidytext	Implementing tidy principles of Hadley Wickham to text mining

Package category	Package Name	Features
Optimization	lpSolve	Interface to Lp_solve to Solve Linear/Integer Programs
Optimization	Minqa	Derivative-free optimization algorithms by quadratic approximation
Optimization	Nloptr	NLopt is a free/open-source library for nonlinear optimization
Optimization	Rglpk	R/GNU Linear Programming Kit Interface

Package category	Package Name	Features
Computer vision	magick	importing / converting to/from all formats / basic image manipulation
Computer vision	imageR	image processing library based on “CImg” (interpolation, resizing, filtering, fourier transformations, denoising, gradients, blurring)
Computer vision	OpenImageR	an image processing toolkit (hashing, edge detection, manipulation)

Disclaimer:

• The contents of this document are to best of my knowledge and based on my own experiences only. Some data and names MAY BE tweaked/masked to take care of data privacy, sensitivity and business sensitivity aspects if applicable. The information provided is purely to highlight experience gathered with clear business impact created and NO WAY RELATES TO ANY ORGANIZATION or ORGANIZATION's OPINIONS, VIEWS.
• Intent is for knowledge sharing and continuous learning as much as possible.
• Focus is also to share from the quorum and leverage from lessons learnt, continuous learning.