Predictive Analytics – Astrology to AI

Introduction

Analytics is divided into 3 categories namely, Descriptive, Predictive, and Prescriptive. Predictive Analytics basically deals with analyzing past data to make predictions about the future with the help of statistical modeling. Some of the main use cases of predictive analytics are to predict

• Demand for products/ services
• Customer churn
• Employee attrition
• Fraudulent transaction
• Loan defaults
• Insurance claims
• Stock market fluctuations

This article tries to address the following points

• Brief History of Predictive Analytics
• Present Scenario
• Process flow of a Predictive Analytics Project
• Business Cases where Predictive Analytics has revolutionized the market.
• Concluding Remarks

A Brief History

Now let’s try to understand how predictive analytics came into the picture. As many people would think it is not of the recent past (50-60 years). Yes, it has been used extensively and in a more systematic way after the invention of computers, but the history goes a long time ago!

Ancient people believed that celestial objects like Sun, moon, stars, and planets always had an astral effect and would affect people and events on Earth. This made many ancient civilizations like Greek, Babylonia, etc. collect very accurate data about the position of these objects in the sky.

A Greek astronomer Hipparchus way back 2000 years ago commented that “It is imperative to have excellent data to make accurate and reliable astrological predictions or forecasts!”. Many Royal Families had an astrologer in their court whom they would consult before making a crucial decision.

Whether we believe in astrology or not is secondary here but what is important to observe is that it created the first foundations for systematic data collection and later used this data to make predictions about the future in other words Predictive Analytics!

Later we also see one of the first-of-its-kind business applications of predictive analytics and modelling. In the year 1689 Lloyd and company in London, which deals with insurance and re-insurance business used data from past ship voyages to determine/predict the risk for different shipping routes and fixed the premium for re-insurance for shipping companies based on the number assigned to the risk.

Present Scenario

Fast forward to the present day. We live in the age of information; we are generating a staggering 2.5 quintillion bytes of data per day! To process this quantum of data would be impossible for Humans. This is where the beauty of computers and programming comes into the picture. With the increasing processing powers of our computers and many open-source programming languages like Python, R, etc., we can process, analyse, and build state-of-the-art Machine Learning Models that are able to make more and more accurate predictions. We are also trying to mimic the human brain in Artificial Neural Networks, all these modern advances fall under the broad category of Artificial Intelligence (AI). There is so much research going on in this field and so much is yet to be known.

Framework for Predictive Analytics Project:

• Problem or Opportunity Identification:

A good analytics project starts with the ability of the organization to define the problem clearly! Domain knowledge and expertise play a crucial role here. This phase is all about asking the right questions.

• Dataset Identification:

The most crucial part of any predictive analytics model building is data.  Predictive analytics is all about analysing the vast amount of data into trends. This collected data helps us with future predictions and remain ahead of the competition. In addition to data available within the organization, we may have to collect the data from external sources.

• Data Pre-processing:

Once we have decided on the dataset, it’s time to start pre-processing the dataset to remove outliers, errors, and discrepancies. From past experiences, it is evident that data processing takes a significant proportion of any analytics project cycle time.

• Model Building:

This is an iterative process that aims to find the best model. Several statistical tools and measures will be used to finalize the model. It is important to split our dataset into multiple training and validation sets to avoid overfitting.

• Model Deployment and Communication:

The primary objective of the analytics is to come up with actionable insights that can be deployed. The communication of model output to clients, especially the top management, which is responsible for making systemic changes, plays a crucial role. Innovative data visualization techniques like dashboards come in handy at this stage. Deployment of the model may involve developing software solutions and products.

Few Applications of Predictive Analytics:

• Amazon – Uses predictive analytics to recommend products to their customers. It is reported that 35% of Amazon’s sales are achieved through their recommender system.
• Netflix – Predicts which movie their customer is likely to watch next. 75% of what customer watch on Netflix is from product recommendations.
• Google –Predicted the spread of H1N1 flu using the query terms.
• Polyphonic HMI – Predicts whether a song will be a hit using Machine Learning Algorithms. Their product ‘Hit Song Science’ uses mathematical and statistical techniques to predict the success of a song on a scale of 1 to 10.
• Capital One Bank – Predicts the most profitable customer.

The list goes further but we can clearly see from the above examples how major companies are using predictive analytics to dominate their business and stay ahead of the competition.

Examples of Industry-wise Predictive analytical problems and data resources.

Why Quation?

Want to know whether the stars will shine for your business in the coming days? Connect with us and talk to our dedicated team of experts. With almost a decade of presence in the industry and has worked with many esteemed clients across myriad sectors spread across the globe, we sure can point the way to the stars!

References

• Siegel E (2013). “Predictive Analytics: The Power to Predict who will Click, Buy, Lie or Die”, John Wiley and Sons, Hoboken, NJ
• Green K (2006) “The $1 Million Netflix Challenge”, MIT Technology Review, October 2006.
• Carneiro H A and Mylonakis E (2009), “Google Trends: A Web-Based Tool for real Time Surveillance of Disease Outbreaks”, 49(10), 1557-1564.
• Annon (2003), “Major Music Labels Use Artificial Intelligence to help determine Hitability of music”, Music Industry News Network, 25 February 2003.
• Davenport T H and Harris J G (2007), “Competing on Analytics – The New Science of Winning”, Harvard Business Publishing School Corporation.
•  U Dinesh Kumar (2017) “Business Analytics – The Science of Data Driven Decision Making”