- You are here:
- Home
- Getting Started
- Understanding Predictive analytics

In this post we will discuss the below topics:

- What is predictive analytics
- Difference with other forms of analytics
- Emphasis on large data sets
- Types of predictive analytics problems
- Challenges in predictive analytics

There are hundreds of areas in life where we need predictions in order to prepare in advance for an uncertain scenario. Simplest example can be weather forecasts that helps us plan in advance about organizing a trip or an outdoor sports event etc.

All such predictions are made on the basis of probability theory. Probability tells us about the likelihood of an event based on past data. In predictive analytics we calculate the probabilities for various outcomes based on historical data to make predictions. So, in effect all that predictive analytics does can be done with a paper and pencil had the data size been limited.

There are four types of analytics: descriptive, diagnostic, predictive and prescriptive. We’ll discuss each one in brief to understand how they differ.

This derives information from the data that helps us quickly understand “what happened” during a given period in the past and verify the effectiveness of a strategy. They can carry out this analysis using certain parameters of their choosing.

If one wants to go deeper into the data that have been collected to understand “Why something happened,” BI tools helps to get such insights. However, it has limited ability to give you actionable insights. It basically provides a very good understanding of a limited piece of the problem you want to solve.

If you can collect contextual data and correlate it with a certain success parameter, you enter a whole new area where you can get a visibility into the future. Essentially, you can predict what will happen to the success parameter if you change certain variables in the data.

Once you get to the point where you can consistently analyze your data to predict what’s going to happen, you are very close to being able to understand what you should do in order to maximize good outcomes and also prevent potentially bad outcomes.

If we try to analyze the development of analytics over the past few years we will find a correlation with the phenomenal rise in collection and storage of data by businesses over these years. As organizations collect more data, there is a natural tendency towards using the data to improve estimates, forecasts, decisions, and ultimately, efficiency. The ways and means to achieve these results is what we have termed as analytics. So the most important difference between conventional statistics and today’s analytics is the quantum of data that they deal with. To this effect predictive analytics draws from several techniques that were in use for decades: pattern recognition, statistics, machine learning, artificial intelligence and data mining.

Predictive analytics problems come in many hues right from predicting about weather to the preference of movies to spam detection, credit score, stock prices, fraud detection, insurance score and so on. If you work in any field where you need to make some kind of forecast, be it marketing, finance or supply chain, chances are you can use some or the other predictive analytics concept to make your job easier.

The primary challenge in predictive analytics is having the data in the right form. At the end, a good analysis can go waste if it’s not based on proper data. Be it predictive analytics or any other type of analytics, collecting and storing the right data in the proper manner is the first step in getting useful insights.