theme: Olive Green, 8 autoscale: true
*Amit Kapoor* [amitkaps.com](http://amitkaps.com) *Anand Chitipothu* [anandology.com](http://anandology.com) *Bargava Subramanian* [bargava.com](http://bargava.com)
- Download the Repo: https://github.com/amitkaps/full-stack-data-science
- Finish installation
- Run
jupyter notebookin the console
data scientist: the people who are building products from data
- Data Management
- Modelling & Prototyping
- Product Design
- Data Engineering
"Jack of all trades, master of none, though oft times better than master of one."
- Data Management: data ingestion & wrangling
- Modelling & Prototyping: statistics, visualisation, machine learning
- Product Design: data narrative, dashboards, applications
- Data Engineering: data pipelines, cloud infrastructure
- Solve a business problem.
- Understand the end-to-end MLaaS approach
- Build a data-driven ML application
- Simple and intuitive
- Go wide vs. go deep
- Practical and scalable
Session 1: Introduction and Concepts
- Approach for building ML products
- Problem definition and dataset
- Build your first ML Model (Part 1)
Session 2: Build a Simple ML Service
- Build your first ML Model (Part 2)
- Concept of ML Service
- Deploy your first ML Service - localhost API
Session 3: Build & Evaluate ML Models
- Feature Engineering
- Build your second ML model
- ML model evaluation (metrics, validation)
Session 4: Practice Session
- Practice problem overview and data
- Build your ML Model
- Build your API
Session 5: Build a Simple Dashboard
- Concept of Dashboard design
- Create your first dashboard
- Integrate ML model API with dashboard
Session 6: Deploy to cloud
- Get started with cloud server setup
- Deploy your ML service as cloud API
- Deploy your dashboard as cloud service
Session 7: Repeatable ML as a Service
- Build data pipelines
- Update model, API and dashboard
- Schedule ML as as Service process
Session 8: Practice Session & Wrap-up
- Deploy on cloud - dashboard and API
- Best practices and challenges in building ML service
- Where to go from here
Two cases / dataset in the Workshop
- Loan Default
- People Attrition
- A start-up providing loans to the consumer
- Running for the last few years
- Now planning to adopt a data-driven lens
What are the type of questions you can ask?
- What is the trend of loan defaults?
- Do older customers have more loan defaults?
- Which customer is likely to have a loan default?
- Why do customers default on their loan?
- Descriptive
- Inquisitive
- Predictive
- Causal
- Descriptive: Understand Pattern, Trends, Outlier
- Inquisitive: Conduct Hypothesis Testing
- Predictive: Make a prediction
- Causal: Establish a causal link
It’s tough to make predictions, especially about the future. -- Yogi Berra
- Human Learning: Make a Judgement
- Machine Programmed: Create explicit Rules
- Machine Learning: Learn from Data
[Machine learning is the] field of study that gives computers the ability to learn without being explicitly programmed. -- Arthur Samuel
Machine learning is the study of computer algorithm that improve automatically through experience -- Tom Mitchell
- A pattern exists
- It cannot be pinned down mathematically
- Have data on it to learn from
"Use a set of observations (data) to uncover an underlying process"
FRAME ——← ACQUIRE ——← REFINE ——←
\
TRANSFORM →——
↑ ↘
| EXPLORE
↓ ↗
MODEL →——
/
INTERACT →—— BUILD →—— DEPLOY →——
- Frame: Problem definition
- Acquire: Data ingestion
- Refine: Data wrangling
- Transform: Feature creation
- Explore: Feature selection
- Model: Model creation & selection
- Deploy: Model deployment
- Build: Application building
- Interact: User interaction
- What are the types of data on which we are learning?
- Can you give example of say measuring temperature?
- Categorical
- Nominal: Burned, Not Burned
- Ordinal: Hot, Warm, Cold
- Continuous
- Interval: 30 °C, 40 °C, 80 °C
- Ratio: 30 K, 40 K, 50 K
- Categorical
- Nominal: = , !=
- Ordinal: =, !=, >, <
- Continuous
- Interval: =, !=, >, <, -, % of diff
- Ratio: =, !=, >, <, -, +, %
Application Attributes
- age: age of the applicant
- income: annual income of the applicant
- year: no. of years of employment
- ownership: type of house owned
- amount : amount of loan requested by the applicant
Behavioural Attributes:
- grade: credit grade of the applicant
Question - whether the applicant will default or not?
default amount grade years ownership income age
------- ------- ------ ------ --------- -------- ---
0 1,000 B 2.00 RENT 19,200 24
1 6,500 A 2.00 MORTGAGE 66,000 28
0 2,400 A 2.00 RENT 60,000 36
0 10,000 C 3.00 RENT 62,000 24
1 4,000 C 2.00 RENT 20,000 28
- Categorical
- Nominal: home owner [rent, own, mortgage]
- Ordinal: credit grade [A > B > C > D > E]
- Continuous
- Interval: approval date [20/04/16, 19/11/15]
- Ratio: loan amount [3000, 10000]
Features:
age,income,years,ownership,grade,amount
Target: default
Training Data: $$ (\mathbf{x}{1}, y{1}), (\mathbf{x}{2}, y{2}) ... (\mathbf{x}{n}, y{n}) $$ - historical records
Given a set of feature
Learning Paradigm: Supervised
- If
$$y$$ is continuous - Regression - If
$$y$$ is categorical - Classification
#Load the libraries and configuration
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
from sklearn import tree
from sklearn.externals import joblib
import firefly#Frame - predict loan default probability
#Acquire - load historical data
df = pd.read_csv("../data/historical_loan.csv")
#Refine - drop NaN values
df.dropna(axis=0, inplace=True)
#Transform - log scale
df['log_age'] = np.log(df.age)
df['log_income'] = np.log(df.income)#Model - build a tree classifier
X = df.loc[:,('age', 'income')]
y = df.loc[:,'default']
clf = tree.DecisionTreeClassifier(max_depth=10).fit(X,y)
joblib.dump(clf, "clf.pkl")
#Build - the model API
%%file simple.py
import numpy as np
from sklearn.externals import joblib
model = joblib.load("clf.pkl")def predict(age, amount):
features = [age, amount]
prob0, prob1 = model.predict_proba([features])[0]
return prob1
#Deploy - the ML API
! firefly simple.predict
#Interact - get predictions using API
simple = firefly.Client("http://127.0.0.1:8000")
simple.predict(age=28, amount=10000)Variables
-
age,income,years,ownership,grade,amount,defaultandinterest -
What are the Features:
$$\mathbf{x}$$ ? -
What are the Target:
$$y$$
Features:
ageincomeyearsownershipgradeamount
Target: default
- Simple! Just read the data from
csvfile
- REMOVE - NAN rows
- IMPUTATION - Replace them with something?
- Mean
- Median
- Fixed Number - Domain Relevant
- High Number (999) - Issue with modelling
- BINNING - Categorical variable and "Missing becomes a category*
- DOMAIN SPECIFIC - Entry error, pipeline, etc.
- What is an outlier?
- Descriptive Plots
- Histogram
- Box-Plot
- Measuring
- Z-score
- Modified Z-score > 3.5 where modified Z-score = 0.6745 * (x - x_median) / MAD
- Single Variable Exploration
- Dual Variable Exploration
- Multi Variable Exploration
Encodings e.g.
- One Hot Encoding
- Label Encoding
Feature Transformation e.g.
- Log Transform
- Sqrt Transform
Types of ML Model
- Linear
- Tree-Based
- Neural Network
Choosing a Model
- Interpretability
- Run-time
- Model complexity
- Scalability
- Easy to interpret
- Little data preparation
- Scales well with data
- White-box model
- Instability – changing variables, altering sequence
- Overfitting
Bagging
- Also called bootstrap aggregation, reduces variance
- Uses decision trees and uses a model averaging approach
Random Forest
- Combines bagging idea and random selection of features.
- Similar to decision trees are constructed – but at each split, a random subset of features is used.
How to choose between competing model?
- Error Metric (Business Decision)
- Hyper-Parameter Tuning
- Cross-Validation
If you torture the data enough, it will confess. -- Ronald Case
- Data Snooping
- Selection Bias
- Survivor Bias
- Omitted Variable Bias
- Black-box model Vs White-Box model
- Adherence to regulations
*Amit Kapoor* [amitkaps.com](http://amitkaps.com) *Anand Chitipothu* [anandology.com](http://anandology.com) *Bargava Subramanian* [bargava.com](http://bargava.com)