A dynamic Markov model for nth-order movement prediction

Prediction of the location and movement of objects is a problem that has seen many solutions put forward based on Markov models. The usual method involves the use of historical data for building a stochastic model in order to make future predictions. Here we present a method for predicting movement of an object using a Markov Model that is not populated by historical data from previous experiments. The proposed method introduces a novel mechanism that dynamically updates the transition probability matrix through analysis of stochastic properties of the data as it is collected. The model gives high accuracy predictions on an object's immediate next movement using a range of orders with results ranging from 79% to 96% dependent upon the type of movement exhibited by the object and order of the model.