Scribd
Upload
2 views18 pages

Comprehensive Guide to Machine Learning

The document outlines various aspects of machine learning, including classification, clustering, and regression, with a focus on practical applications and model evaluation. It covers specific models such as Naïve Bayes, Random Forest, and Support Vector Machine, as well as techniques like hyperparameter tuning and feature engineering. Real-life examples, such as predicting stock movements and identifying top songs, illustrate the expected tasks of data scientists in applying these methods.

Uploaded by

Avanish Kumar Singh
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
2 views18 pages

Comprehensive Guide to Machine Learning

The document outlines various aspects of machine learning, including classification, clustering, and regression, with a focus on practical applications and model evaluation. It covers specific models such as Naïve Bayes, Random Forest, and Support Vector Machine, as well as techniques like hyperparameter tuning and feature engineering. Real-life examples, such as predicting stock movements and identifying top songs, illustrate the expected tasks of data scientists in applying these methods.

Uploaded by

Avanish Kumar Singh
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

Machine Learning

Dimensions
• Statisticians
• Data Science Theorists
• Business Applications
• Functional Managers
• Model Users
Coverage
• End to End Machine Learning Problem- Classification
• With clean and structured data
• End to End Machine Learning Problem- Classification
• With missing data and unstructured outliers
• Cross Validation
• Feature Engineering
• Hyper Parameters Tuning
• ROC and AUC
• End to End Machine Learning Problem- Clustering
• Basic Clustering
• Using non numeric data
• Combining Clustering with Classification (Ensemble)
• End to End Machine Learning Problem- Regression
• Machine Learning as a Service
• Azure Machine Learning Studio
End to End Machine Learning
• Models Covered
• Naïve Bayes
• QDA
• Logistic Regression
• Classification- RPART
• Random Forest
• Support Vector Machine
• Gradient Boosting
• XGB
• kNN
• Basic Process
• Use each of this method
• Develop an evaluation matrix
• Compare the models and choose the most appropriate
Consolidated Machine Learning (IBP)
Integrated View of ML
• Training Task
Make Task • Testing Task

• Any of the Cluster, Classif, Regr


Make a Learner • Define Predict Method

Train Learner
for Training Task

Predict on Test
Task

Evaluate
Practical Hands on
• Gestational Diabetes
• Dataset Description
Describing Confusion Matrix
Prior Probability
Good Customers 20 20/30 0.667
Bad Customers 10 10/30 0.333
OPMD- Naïve Bayes Total Customers 30

Define Problem

Likelihood of the new customer given good: 3


20
1
Likelihood of the new customer given bad:
10
3 20
Posterior Probability that customer is good: ∗ =0.1
20 30
Posterior Probability that customer is bad: 1 10
∗ =0.03
10 30
StatQuest with Josh Starmer

OPMD- Discriminant Analysis


OPMD- Logistic Regression
OPMD- Classification
OPMD- Random Forest
OPMD- SVM
Tuning parameters: Type, Kernel and gamma
OPMD- GB and XGB
OPMD- kNN
Real Life Examples of Machine
Learning-1
• Identifying Top 10 Songs
• Dataset Description
• What a Data Scientist is Expected to do?
• General: Build model with Train dataset and predict which songs would come
Top 10 from the Test dataset. Assess the accuracy.
• Specific:
• Which un-tuned algorithm works the best?- (Try out all)
• Tune any one of the algorithm and see the performance.
Real Life Examples of Machine
Learning-1
• Stock Movement Predictions
• Dataset Description
• What a Data Scientist is Expected to do?
• Script is provided for the following activities:
• Basic un-tuned algorithms
• Clustering the stocks with k=3
• Splitting the original dataset into 3 clusters.
• A trial calculations of model improvement with RandomForest and SVM. Since the
improvements are observed, you have to do the following:
• Re-build all the models with Cluster 1 as data. Which models show
improvement and which do not? Why?

You might also like