ITC Analytics Accelerator

Workshop

• Forecast vs Prediction
• Logistic Regression | Out put is a probability
• OKR <- Empathy as a theme

Model Performance Testing / Model evaluation

Model evaluation should be part of the data driven project matrix pipeline

• Confusion Matrix
  • To Identify false positives & true negatives
  • Precision = True positives /(True positives + False Positives) |
  • Recall = True positives / (True positives + False Negatives) | In a case -> Identify the person with heart disease
  • High sensitivity -> High true positives
    • High sensitivity = True positives /(True positives + False Negatives)
  • High specificity -> High true negatives
    • High sensitivity = True Negatives /(True Negatives+ False positives)

Code

General

* import pandas as pd
* var = pd.read_csv('used_cars_maruti.csv')
* var.describe() | statistics of the data
* var.info() | summary
* var.shape | Rows and columns
* var[0:5] | range of rows to be shown. Minus 5 is also possible
* var['Column Title'][0:5] | First 5 rows are shown
* var[['Column Title','Column Title']][0:5]
* header = list(data.columns.values) | capture the column names (header)
* var.sample(3) | Random sampling of any 3
* var.describe(include = 'all')
* newvar = var [var.Column Title < 15] | Filtering
* var.sort_values('Price', ascending = False) [0:5]
* vars.head() | vars.tail()
* corr_mat = cars_df[['Price', 'Age', 'Power','Mileage']].corr() | Correlation
** corr_mat

Plotting =

* import matplotlib.pyplot as plt
* import seaborn as sn


# jointplot
sn.jointplot(x='Price',y='Age',
data=cars_df,kind='hex');
plt.show()


# Heatmap
sn.clustermap(corr_mat, cmap="GnBu");
plt.show()


# Heatmap with colour
sn.clustermap(corr_mat,cmap='coolwarm',
   standard_scale=1);
plt.show()

# Heatmap with correlation | data should be fully numbers
plt.figure(figsize=(9,5))
sn.heatmap(cars_num.corr(),annot=True, linewidth=0.5)

OVERVIEW

5i: 
Ideation: defining
Intelligence: process of prediciton
Inception: model building
Intervention: Developing a segment
Independance: Live


DATA ENGINEERING
collection
Sructuring
Cleaning


ALGORITHUMS
Classification & Regression
Decision Tree
Tree based Algorithums (RF model
Ensemble models


MODEL PERFORMANCE
Model selection

AI & ML
ML
supervised
Unsupervised

DEEP LEARNING
Reinforment learning
CNN

IMPLEMENTATION of ADVANCED ANALTICS
Tracking

CHANGE MANAGEMENT
Piller & Practics

Advanced Analytics

Value of AA

• Evolution: Excel > Storage > Tableu > Alerts > predicting the future > ML >
• Applications:
  • Descriptive (historical. advanced statistics) >
  • Predictive (Future. Uses ML. detection of fraud) >
  • Prescriptive (influence the future. Data driven decision making); Optimization;

Five concepts

Statistical Learning

• Frame of understanding the data based on statistics
• Eg: L.Regression: Used in predictive analytics
• Linear model: regression or classification

Inference vs prediction

• Inference : process of evaluating the relationship of predictor and the response
• Prediction: process of using a model that is yet to happen

Flexibility vs Interpretability of models

• Flexibility: capture a wide range of phenomena (Data fit)
• Interpretability : Design a Pattern to facilitate the decision

Regression vs classification

• Regression : Continuous variable.
• classification: discrete variable

Supervised vs unsupervised learning

• Supervised " Labelled data sets, Training , CNN , Decision tree
• unsupervised : Classification
  • Anomaly detection
  • Find hidden structure
  • Eg: Customer segmentation; Purchase patterns, simplify data, monitoring