Search

In order to improve performance of personalized label recommendation, the recommendation algorithm of implied relation topic model (RTM) for microblog personalized label based on Gibbs sampling inference is proposed. Firstly, imaging form is used to express potential local information in microblog and to conduct top-k similar user discovery for users represented as user topic distribution, and then the frequency of all labels in these users is calculated to recommend the label mostly related to the users. Secondly, to dig potential topic information, enhancement cosine similarity RTM model with penalty term is used to name the microblog label, which greatly improves the influence of joint modeling on potential topic generation label, and the relationship between overall label and topic can be found; finally, it can be seen from real experimental result that the proposed recommendation method is superior to the selected TF–IDF, RTMSA and other classic label recommendation algorithm, so as to verify effectiveness of the algorithm.

An approximation algorithm for high-dimensional, continuous-state stochastic dynamic programming was first presented based on an orthogonal array (OA) state space discretization and a Multivariate Adaptive Regression Splines (MARS) value function approximation. Given the popularity of Number Theoretic Methods (NTM), this paper compares OA-based experimental designs and NTMs for state space discretization using a ninedimensional inventory forecasting problem. The statistical model employed for future value function approximation is Artificial Neural Networks (ANN).