The Evening Brew !

Evening brew session 05/04/2018 - TIQRI

Learn something collaboratively - Machine Learning

Machine learning Building a model from example inputs to make data driven prediction vs. following strictly static program instructions. Mainly system learns how to solve a problem from example data rather than following a specific logic written with branching statements.

In traditional programming

We carefully analyses the problem and write code, this program then reads data then uses control logic we wrote carefully and determines correct part of the program to execute, which eventually produce end results.

Traditional control logic

If, switch case, for, while, etc…

In Machine learning

We gather a dataset from a problem domain, and pre-process it in to a form where machine learning algorithm can understand. Then we feed the data to an algorithm which analyses the data and produce a model . This model implements the solution to resolve the problem based on the data. the initial data we used drives the prediction logic, not a logic written by a programmer.

How can an Algorithm learn it self ?

check the video

We will be useing Azure ML Studio to get-in to basics

https://studio.azureml.net/

Some example flow diagram,

Why ML studio

Azure Machine Learning Studio is a collaborative, drag-and-drop tool you can use to build, test, and deploy predictive analytics solutions on your data. Tutorials, videos, and example models show you how to use Studio to build and deploy machine learning models.

We will be focusing mainly on machine learning flow, and the algorithm selection

What is 'Machine learning flow' means ?

Asking the right question > Preparing data > Selecting the algorithm > training the model > testing the model

check the video

We have prepared some data sets here to begin with, Each team will have to go through each data set which they needs to analyse using ML studio, and present findings to the crowed with appropriate demonstration.

1. Diabetes-dataset This dataset is originally from the National Institute of Diabetes and Digestive and Kidney Diseases. The objective of the dataset is to diagnostically predict whether or not a patient has diabetes, based on certain diagnostic measurements included in the dataset. Several constraints were placed on the selection of these instances from a larger database. In particular, all patients here are females at least 21 years old of Pima Indian heritage. More

2. Function-approximation-dataset-1 This is some common trignometry function :)

3. Function-approximation-dataset-2 This is S = ut + 1/2* at^2

4. kickstarter-projects-success-dataset Data are collected from Kickstarter Platform try to predict success/failure out come of a project. More

5. interview-attendence-dataset The data pertains to the recruitment industry in India for the years 2014-2016 and deals with candidate interview attendance for various clients. The details are largely self explanatory. More

Try few algorithms to determine outcome of each data set,

At the end of the session

Try to figure out diffrent types of evaluation

How do I ?

Begin

Azure Machine Learning Studio Documentation

Find data sets

https://www.kaggle.com This is only one in many good chioces... why don't you try google!

Algorithm Selection

Machine learning algorithm cheat sheet

How to choose algorithms for Microsoft Azure Machine Learning

Capabilities

Overview diagram of Azure Machine Learning Studio capabilities

Try some serious s#!t

Create text analytics models in Azure Machine Learning Studio