The real challenge with computational decision making is to decide when the model is so wrong that it is no longer useful. Usual problems with techniques of computational decision making include assumptions regarding the nature of variables (e.g. Linearity, Stationarity, and Normality) as well as unavoidable statistical biases such as over and / or under fitting, omitted variable biases, and sampling biases. Also decisions are often constrained into feasible and infeasible solutions depending on business domain and business considerations and many computational decision making techniques struggle in the presence of boundary, equality, and inequality constraints.
At MarkeTopper Securities Pvt. Ltd we have tried to answer the challenge by developing a host of proprietary trading frameworks. These are designed to allow for on-demand scalability with trusted stability. The frameworks are purposely built to support MarkeTopper’s varied business operations ranging from development of artificial intelligence driven trading algorithms to their automated execution.
Artificial intelligence can be understood as the ability of a computer to make decisions either optimal or acceptable. Decision-making in this context is a process of selection of a particular course of action among several alternative possibilities. When this decision making is automated using computational techniques such as neural networks and decision trees it is called computational decision making. Just like how Deep Blue (deep blue was the program by IBM to teach computers to play and beat the best chess players) had to decide what moves to make against Garry Kasparov, computational finance (and finance in general) is also all about making intelligent decisions. There are many fields dedicated to trying to find optimal ways of making such decisions including, but not limited to, statistics, operations research, and artificial intelligence.
At MarkeTopper Securities Pvt Ltd, we are engaged in Computational Decision Making with many of these techniques, ranging from Technical Analysis to Probabilistic Belief Networks, which are powered by 400 Teraflops of High Throughput Computing Infrastructure.