How to predict the future? Knowledge consilience of physics and economics
- Random selection (Stupid but statistically average)
- General solutions
- Machine (linear method)
- Three drawbacks: focusing on recent issues, seeing what is interesting, assuming local tendency to global
- Expert (or Boss)
- Newly emerging solutions
- Knowledge based method
- Complex science (Chaos & Fractal)
Today, I wanna talk about how to predict the future as an example of knowledge consilience between physics and economics.
Conventionally, we consider machine based, human based and expert based (or boss based) methods for the future prediction. However, all of those methods have fundamental limitations.
As an application of the future prediction, we consider the stock investment. If we can know the 10 year later price of certain stocks, we can earn the money and our investment will be more safe and stable. However, it is almost impossible to predict the future, that is, the future price of any stock. Then, it can be a question how we can select appropriate stocks for investment even if we can not know the future. Some says that we will invest based on our inherent feeling but based on experimental human heuristic decisions will be usually failed. The probability to select stocks which help us to earn the money is much less than 50%, or extremely less than 10%. Then, how can we enhance the correctness of our prediction.
If the prices of one stock is 8, 9, 10 in 2008, 2009, 2010, respectively. What will the price of the stock be in 2011, most probably? Even if there are also many another facts to control the price except the previous prices, we can say that the price will be 11, most probably. This future prediction method is a linear approach. However,even if the linear methods are mathematically simple, we don't know whether it is the perfect solution or not. Recently, French scientists prove that the natural human could not think the situation linearly. In stead, the natural people prefer a logarithm approach. Therefore, we can conclude that in order to find more accurate prediction, we need to use higher order and more complex methods than linear method. The linear method is actually a representative method of machine approaches.
How about the performance of the human method? Is it really better than the machine method for the prediction of the future? Recent economists found that the human heuristic approaches also have many significant drawbacks. Even if many people say that the feeling of yourself is better than the calculation based on the machine method, we must remind that heuristic approaches also has limitations. Human usually focus on recently emerging issues even if it is not most significant for the whole event, is interested in what he want to know, and assume local results to be global conclusion. Hence, the prediction resulted by human heuristic decision is not the only solution to replace the machine approach.
Lastly, how about expert based methods? It is already proved that the performance of the future prediction of the crowd is clearly better than the expert's prediction results once the number of the crowd is not too small.
Since a boss can be considered as an expert, it is obvious that the future prediction performance of a boss is worse than that of the crowd if there is a way to collect the opinion of the crowd efficiently.
In summary, we discussed that the performance currently well using approaches for the prediction of the future price of stocks are not good. Machine methods, human methods, expert (or boss) methods will bring all limited performance. Hence, it is really wonder whether there is more advanced methods to predict the future. Notice that we assume that the future can be predicted by the recently obtained data such as history based on the successful linear prediction history in digital audio, video and communication areas.
I'd like to introduce newly emerging methods as advanced future prediction approaches useful for the prediction of the stock price. The emerging methods are first, knowledge based approach, second, complex science (chaos & fractal) approach, and third, bio-mimicry approach. We will discuss how these new methods to enhance the performance of the future prediction later on.