Search

Hierarchical Temporal Memory (HTM) is an advanced machine learning technique that aims to capture the structural and algorithmic properties of the neocortex. It is an online machine learning method and can make predictions and classifications. This paper is an application of the HTM theory. We use the HTM theory to diagnose diseases. First, we encode the cases of the patients into the format that can be recognized by the HTM. Second, we train the HTM by using the encoded cases. At last, we can predict the disease of a patient according to the symptoms. We do experiments on real nephrosis diagnostic datasets. The experimental results show that we can predict the diseases accurately and efficiently.

Skyline is defined as a set of objects in a multidimensional dataset. It returns objects which are not dominated by any other objects in the set. An object p dominates object p' if and only if it is not worse than p' on all of the attributes (dimensions) and is better than p' on at least one attribute. Given the same kind of dataset, the skyline group query returns groups which are not dominated by any other groups in the set and each group has the same number of objects. Although the skyline group query has been investigated in recent years, most techniques are designed for static datasets. However, data are changing with time in many practical applications nowadays and query processing techniques for dynamic datasets are not available. Therefore, finding skyline groups on data streams are highly required. In this paper, we propose a new algorithm to find the skyline groups on data streams. We use two data structures CL and GSM to store the immediate results. CL stores the candidate objects that may become members of the skyline groups. GSM stores the immediate results of the dynamic programming. We do experiments on synthetic datasets. The experimental results show that the algorithms proposed can find skyline groups efficiently.

According to the data feature of customer's consumption records of bank POS machine and the analysis depending on the actual requirements, a new modeling framework on consumption behavior of bank POS machines is presented in this paper, and further research on the implementation method of main aspects in the model is carried out. Firstly, we conduct data discretization and customer segmentation by K-means algorithm and Kohonen network clustering algorithm respectively, analyze and compare the results comprehensively, and ultimately get the optimum result of extracting high quality customer. Then in the process of mining the high quality customer's consumption characteristic, we use C5.0 algorithm to analyze the experiment and evaluate the experimental results after getting the high quality customer's consumption records through customer segmentation results. Finally we get the knowledge base of high quality consumer's consumption behavior which can give support to banks and merchants to make a favorable decision.