An Improved Apriori Algorithm Based on LinkedList for the Prevention of Clinic Pharmaceutical Conflict

Abstract

Investigation of association relation for medical prescription appears as an essential part for the treatment of patient in clinical pharmacy. Not often, however occasionally, the pharmaceutical conflict that could possibly relate to unpredictable hazard to patient may occur. Therefore, clinical physicians apply specific software package embedded in HIS (Hospital Information System) to sensitively and instantly discover the potential danger from the electronic prescription. In some HIS systems or independent software packages, Apriori algorithm is applied in pharmaceutical conflict checking, however, the efficacy is encumbered due to the mechanism of the algorithm. The mentioned algorithm Apriori is a classical algorithm for association rules mining, it applies an iteration method called searching step by step, to explore K itemset by (K-1) itemset. Each time when it explores a K-itemset, Apriori algorithm must scan the whole database, which causes the obvious reduction of efficiency due to the requirement of constant database scanning. To solve this problem, an improved Apriori algorithm based on Linkedlist is applied. Due to the decrease of transactions number, the operating efficiency of the algorithm is enhanced during the process of K-itemsets exploration. In order to reduce the number of transactions, the database is transformed into a series LinkedLists which could allocate an element of L1 as header node. After being transformed, the whole database scanning could be skipped. Instead, the corresponding LinkedList is scanned to explore K-itemsets. The proposed method could filter the unrelated transactions during the generation of frequent itemsets. The experiment proves that the new algorithm could release better performance rather than original Apriori algorithm.