FINDING CONSTRAINED SEQUENTIAL PATTERNS FROM MEDICAL DATASETS

Abstract

The problem of mining sequential patterns from supermarket data is an interesting data mining problem and has received a lot of attention to researchers. Mining sequential patterns from supermarket data is actually discovering causal relationship between different itemsets or symptoms that are presents on the patients. Supermarket data contain the records about the information of the itemsets or symptoms of patients besides the patient’s personal information and is ordered in accordance with the time / date of consultation of the patients in the hospital. Such data may offer us the precious information related to the cause and effect one itemset on another on the human body. Although, it gives us the ordering of the occurrence itemsets in the human body, it does not provide us the information about the time intervals within which the successive itemsets may occur in a human body. That is, we don’t know after how much time the next itemset may occur in human body. In this paper, we try address the issue in detail and propose a method of extracting such sequential patterns from such medical data which occurs within a user-specified time intervals. The efficacy of our method is established by experiment conducted with a dataset collected from a private hospital of Saudi Arabia