Machine Learning, AI & Deep Learning Certification
Machine learning methods are used for data analysis, this is where they are similar to data mining, but the main goal of machine learning is to automate decision models. Algorithms are the heart and soul of machine learning and they help computers to find hidden insights.
So in essence machine learning algorithms need to learn. The machine needs to learn from data. Data will have multi dimensions- Type (quantitative or qualitative), amount (big or small size) and number of variables available to solve a problem. Learning algorithms should also be as general purpose as possible. We should be looking for algorithms that can be easily applied to a broad class of learning problems.
The data scientists are responsible for machine learning and getting outputs but the business people are the ones who are going to use it for business purpose so the rules and insights extracted from machine learning should be interpretable. So the output produced by the machine has to be understood by humans, who may not be from the machine learning area.
What differentiates this course are the following:
- Case study based approach, where participants are immersed in the problem.
- Complete package of AI (deep learning and NLP) and machine learning.
The training aims at providing the participants with latest and general purpose machine learning algorithms. At the same time the training aims to deliver some common threads or a common knowledge base which can be used in future for learning a wide range of algorithms.
Large and comprehensive use cases:
- Financial Industry- Participants would learn “Credit Scoring.”
- Health care industry- Use tensor flow for extracting information.
- The Canada police differential treatment story.
- How to find correct weights for food items and have a winning compensation program.
- Predicting customer churn in a telecom industry.
Other than this working use cases on:
- How to reduce a large number of variables.
- Predicting heart attack.
- Convert categorical information to continuous information.
What are the components of the machine learning area
- Share resources for future and current learning
- What kinds of problems can be solved by machine learning
- Appreciate different kinds of data
Cover basic of R to make participants comfortable on the tool
R is the primary tool for data manipulation and machine learning algorithms.
Participants need to be clear with basic statistical concepts such as probability distributions, hypothesis testing
Bring all participants to a level where they are comfortable with statistics which is a mandatory component of machine learning.
Algos- MLR, Logistics and nonlinear regression
Make participants hands on with predicting with regression models
Algos- SVM, decision trees, boosted decision trees, Naïve bayes
- Classification is the most used class of algos in real business.
- Participants should be able to choose the correct algo and use it.
Concepts of ROC, hit rate, kappa statistics and K-S statistics
Participants would be able to know how good a classification model has been fitted.
How to select useful variables out of substantial number of variables
- Learn feature selection methods for regression- Ridge and LASSO
- Feature selection methods for classification methods- Information value based, filter based and wrapper based.
How consumers make decisions and value attributes
How to know the consumer is about to leave you
Hidden Markov Models for churn analysis.
How to use different neural nets for Deep learning
- Boltzman machines
- Convolution networks
- Recurrent neural networks
Different components of NLP
- Parts of Speech
- Text Similarity
Consolidating the learnings