Prediction brings insight into unknown. Accurate predictions can potentially transform businesses, industries, and almost any organization. Marketing, financial services, insurance, retail, and healthcare are just a few industries seeking for accurate predictions to enhance their decisions. Today, humanity more than ever seeks accurate predictions to better react to the climate changes. Due to uncertainties, complexity of the prediction functions and high computation costs, the conventional mathematical modeling approach cannot provide any reliable prediction model. Instead predictive analytics, emerged from data science, identifies patterns in big data to build predictive model for organizations. Although prediction modeling is meant to empower the decision-support systems, in reality it could not be beneficial as expected, as yet identifying the best possible response to the valid prediction is the actual problem that organizations facing now. In fact optimizing the optimal decisions and anticipation of every decision and its consequences must be also predicted and optimized. To scale to the complexity created Dr. Mosavi coins the term “predictive-decision model” a novel integration of prediction analytics with decision modeling, where predictions are optimized and decisions are predicted. Further an intelligent agent makes automated decisions relying on learning algorithms and the decision preferences. This will revolutionize the way decision-support systems function today. (Dr. Amir Mosavi, 2014)
Most prediction exercises are for decision support. And by prediction, we mean using hindsight to gain foresight.Most systems are very complex to realise analytically as such scientists leverage on historical data for data-driven modelling by applying learning algorithms on this data. However, the model, which could be a mathematical model, tree structure and so on, can be integrated into applications to enable the users to be supported while taking a decision as the case may be. This is, in my opinion, what is meant by predictive decision making. The applications of predictive decision model cut across all imaginable domains. The common ground to developing any predictive model despite the domain is data. In addition to data, knowing the right algorithm to use and how to make it perform well on your data are equally important.
In complex systems (or markets) characterised by uncertainties and dynamic competitions, I would suggest an alternative approach to what you are describing. Instead of focusing on getting a "precise" or "accurate" prediction, which is perhaps verging on impossible, you may consider examining the the feasible outcomes under different scenarios. The task of decision-makers is to craft strategies premised on strategic and operational flexibility, which is possible when the pathways towards the feasible outcomes are well understood. You may use real options reasoning to evaluate how the value of your firm (or strategy) would change when market conditions change. Surely, the ability to shift course when circumstances change has in itself a great strategic value.
When operating under dynamic competitive markets, option games reasoning becomes useful in understanding how the firm's competitors' actions could impact their firm's value or the outcomes of their chosen strategy.
In both instances, you are likely to gain insights that a "precise" or "accurate" prediction of outcomes would often miss. You may consider going through my papers that you can download from my page in ResearchGate, published by the Journal of Applied Corporate Finance.
https://www.researchgate.net/publication/274731802_Renewable_Energy_with_Volatile_Prices_Why_NPV_Fails_to_Tell_the_Whole_Story
https://www.researchgate.net/project/Book-Energy-investments-An-adaptive-approach-to-profiting-from-uncertainties
Article Renewable Energy with Volatile Prices: Why NPV Fails to Tell...
Predictions, as already mentioned above, are most of the time done to support decisions under uncertainty(Deciders are incompletely informed about the situation) .
Intuitively, prediction itself is all about trying to determine the state of a system in advance. The system can be either stochastic or deterministic.
Deterministic system means that we can precisely predict the behavior of the system through a well-defined model, Example: Trajectory of mobile in abstract world for simple motion(rectilinear,..)
Stochastic system means that the behavior of the system appears to be random and cannot consequently be predicted precisely through a model. We can only approach a modeling of the system and then a prediction. In this case, Statistics and Probabilities play a crucial role. Some of the most used statistical and probabilistic models are:
-Regression
-Decision Trees
-Bayesian networks
- Markov's models
....
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence