Puneet Mathur's Blog

How to Predict correctly

Prediction as we have seen in my previous blog post requires experiential knowledge. This makes prediction all the more difficult. I would say prediction is an art and science both. Prediction becomes a science if we use techniques which can show cause and effect relationships between different forces which we may hypothesize to predict. Any art form has an element of uncertainity and requires skill to practice. This makes prediction even more difficult. In my 17 years of experience in the field of Prediction be it Corporate prediction or at a personal level helping individuals make decisions which are best for them, I have been able to make out what works and what does not work in prediction. So it is important to know how to Predict correctly. Personal situations are important for the individual however corporate situations are even more complex and require advanced techniques of prediction. Even more complex than corporate prediction is the mundane prediction or predicting for events which effect a whole nation, region, state or economy. One such example is predicting outcome of US Presidential elections 2016 through my book Predictology. In this book, I have used data science to unearth hitherto hidden secret algorithms by using data mining of Astrological datasets of United States of America. The key to predicting correctly can be summarized below: 1. Know your subject of prediction. If you want to predict you must know the object of your prediction fairly well. For example: In my book Predictology I have mentioned that if you do not know your subject you are likely to fail. Not knowing your subject of prediction can completely destroy the outcome of your predictions. The key thing is you should become a master of the subject you wish to predict otherwise you are setting yourself up for failure. 2. Look at your data needs. Ask yourself do I need any data to make this prediction? What is going to be the data source? Is the data source dependable? What is the accuracy rate of data that I am getting? All these questions should go through your mind if you are going to embark on corporate or mundane predictions. It is said Garbage in Garbage Out. Which means if you have garbage for data to feed into your prediction model then what you can expect is garbage or completely error prone prediction. Of course for making a personal prediction you may not need massive amounts of data. 3. Keep the end goal in mind and keep yourself reminded of this goal when you are gathering data cleaning and munging it. Why I stress on this point of keeping to remind yourself of your goal is because I have seen even experienced data scientist teams getting lost in the world of big data. By nature, big data is huge and to lose track of why we started in the first place and where are we from our target is all very important in prediction. This is especially so because every prediction corporate or mundane project would typically last anything between 3 to 12 months and for everyone to stay focused on the goal of the prediction team it becomes extremely important. In my 6 months Flight delay prediction project of predicting the results of a flight. It was very easy to get lost in the data, sometimes

I got so many different insights during data mining and analysis that I thought of starting parallel research into what new pattern discovery was coming forward. But I kept reminding myself and bringing myself back to the original goal of predicting election results. We have looked at some of the key things that enable everybody on the path of prediction to making a good prediction. Do remember to share this blog post if you like it and if you have any questions or comments do post them in the box below.

What is Prediction

Prediction, as defined by Wikipedia, is a statement about an uncertain event it is often but not always, based on experience or knowledge. So prediction always involves an uncertain event in the future. It also requires experience or knowledge. So the key elements which define prediction and forecast involve experiential knowledge about what is going to happen. There are people who say what is going to happen is not known but prediction involves knowing this unknown which is going to happen at a distant point in time. I make another distinction which normally other authors do not make about prediction is that it involves predicting outliers. Traditional Predictive Analytics techniques use statistical methods and are based on the notion that "Past will follow through the future". This means if we see a trend in the past this assumption warrants that this will definitely happen in the future. Ask yourself this question how many events which happen in our lives in our states in our regions and in our countries have happened in the past have also happened in the future. How many 9/11 type of events have happened in the past?, how many Lehman Brothers type of financial debacles happened in the past? How many World wars of the same type happened in the past? The answer to all the above questions is None. These events as per Predictive Analytics are known as Outliers and Random events part of the Standard Error which cannot be predicted. I found in my researches the secret formulae which helps us unearth the models which can predict event Outliers. You can read all about it in my new Book Predictology. Where I have used both Data Science and Astrology to build a model for predicting US Presidential elections 2016. More on this in my next blog posts. Bye for now. Do remember to post comments below and ask any questions that may occur to you.