Introduction to Machine Learning

On this page I am going to share some important concepts, implementation of code for the better understanding of Machine Learning. We will take a look at some of the machine learning models, how we can create them and use them in various scenarios.

Dig: Apply Machine Learning models to gain insights from the Data

As shown in the above image one can get insights from the prepared data using data analysis techniques. Normally we can use Machine Learning techniques to analyse the data. Machine Learning models with relevant and sufficient data can be used to get some really useful insights that any organisation can use for their intended purpose that can be related to problem solving, decision making or profit making, etc. Also as a Data Scientists or Analyst after giving insights we can recommend which new/additional data that organisation should/need to capture to get more useful insights.

Machine Learning Basics

1. Brief Introduction to Machine Learning and Free good sources to learn more about Machine Learning

2. What is Vectorization? Why it’s important?

3. Broadcasting in Python. Related problem with it. Assertion statement to avoid strange bugs.

4. Before starting to machine learning project please take a look at these points.

Machine Learning Models

Prediction Model

In the Machine Learning and Deep Learning we use prediction models to make predictions (for example linear regression, logistic regression, Random Forests, Decision Trees, Neural Networks etc). The definition of Predictive modeling is a process that uses data mining and probability to forecast outcomes. Each model is made up of a number of predictors, which are variables that are likely to influence future results. Once data has been collected for relevant predictors, a statistical model is formulated.

Application of Prediction models ranges from the e-commerce to medical domains. Prediction can be used in various domains for example in e-commerce websites you can predict which visitor can make purchase (that can be converted to buyer), in case of medical images we can predict certain medical conditions, identify organs etc, and so on. The applications of prediction models with good accuracy are limitless :D. It is up-to you as a data scientists or analyst how to make prediction model very successful and practical to use it solve the problems that businesses and societies are facing. For that you should be able to

1. get thorough understanding of the problem that to be solved,

2. get the deeper understanding of the domain knowledge specifically related to the problem

3. identify which features should be used as the inputs for the model

4. make use of derived features (if needed) along with other features to increase the accuracy,

5. most importantly having understanding of the limitations of machine learning models and data that can be captured with the current infrastructure. (here you can not be dreamy about the success be practical and honest about the models)

Now I will show you some implementation of Machine Learning Prediction models in the series of articles:

1. Simple prediction model to identify certain things in an image e.g. identify specific animal or digit in an image (coming soon)

2. End to End prediction model to predict specific characteristics of a user or customer or a person based on his past activities related to specific area of interest (for example on e-commerce website predict whether a visitor to the website is going to buy or not or on a gaming app based on gamer’s activities predict he/she will buy products or not)

Recommendation Engine

1. Simple Recommendation Engine models for e-commerce website to make user specific recommendations. (coming soon)

2. Implementation of Recommendation Engine for user specific recommendation of products to increase the revenue and user satisfaction (coming soon)

3. Recommend products which are frequently sold together applicable to e-commerce websites, malls, shops, Restaurants, Apps etc (coming soon)

Time series:

1. Brief introduction to time series, components (trend, seasonal, irregular) of time series and models used to decompose time series (additive, multiplicative, etc)

2. Data imputing using interpolate, fillna – Time series example

3. Introduction to Facebook prophet. (Coming soon)

4. For a time series data Forecast next month sell using prophet. (Coming soon)

5. Sharing these useful articles on Time Series:
   

6. Anomaly Detection with Time Series Forecasting

7. 3 facts about time series forecasting that surprise experienced machine learning practitioners

Data Visualisation tool:

1. Exploring features of Microsoft Power BI

2. Implementing Prediction Model (Logistic Regression) using Python in Power BI

3. Access the files created during the prediction model implementation on Power BI using Jupyter notebook. Accessing the output table of Prediction model and Calculating Accuracy.

Machine Learning Model Deployment

1. Machine Learning Model Deployment using Flask (Coming soon)

Machine Learning pipelines

1. Introduction to Apache Airflow

2. Apache Airflow installation steps on Ubuntu.

3. Apache Airflow single node and multi-node Architecture

4. End to End Scheduling and running steps from Data Fetching to Machine Learning model train, testing and saving using Apache Airflow. (coming soon)

5. Sharing these interesting video and article:
   

6. Youtube Video – Industrial Machine Learning Pipelines with Python & Airflow by Alejandro Saucedo

7. Article – Problems faced by Bluecore Engineering with Apache Airflow operators and how they solve it using Kubernets operators

Some interesting read:

1. I had no idea how to build a Machine Learning Pipeline. But here’s what I figured

2. Overview of the different approaches to putting Machine Learning (ML) models in production