A Bayesian assumption based forecasting probability distribution model for small samples

In this work, a novel forecasting probability distribution model is presented. Probability distribution plays a role in the function of probability values. Therefore, forecasting the probability distribution function is a challenging process. To that end, the method described in this work loosens the control conditions of the given data set. Subsequently, statistical methods can be applied to the resulting sample data. The distribution functions are then fitted using the cubic spline interpolation method. In this work, the naive Bayes and the Bayesian network methods are adjusted to handle the small sample problem. In addition, the maximal extension clusters are used to determine the conditional function. Two data sets from the UCI repository and a custom data set are used to validate the forecasting model. The experiments show the proposed method can generate an accurate distribution function.