CAJ | 학술논문

针对传统的基于项目的协同过滤推荐算法中数据稀疏问题，以及受时间效应影响推荐准确度较低问题，提出将隐马尔科夫模型与传统的基于项目的协同过滤推荐算法相融合的推荐算法HMM-ItemCF。算法通过隐马尔科夫模型对系统中所有用户的评分行为，与目标用户的历史评分行为进行统筹分析，找到一批用户下一时刻概率最高的评分对象，并将这些评分对象发生概率与传统的项目相似度计算方法相加权得到新的相似度，最终产生推荐结果。仿真实验中对算法的重要参数进行训练，并与其他算法进行对比，证明改进后的算法是有效的。
침대전통적기우항목적협동과려추천산법중수거희소문제，이급수시간효응영향추천준학도교저문제，제출장은마이과부모형여전통적기우항목적협동과려추천산법상융합적추천산법HMM-ItemCF。산법통과은마이과부모형대계통중소유용호적평분행위，여목표용호적역사평분행위진행통주분석，조도일비용호하일시각개솔최고적평분대상，병장저사평분대상발생개솔여전통적항목상사도계산방법상가권득도신적상사도，최종산생추천결과。방진실험중대산법적중요삼수진행훈련，병여기타산법진행대비，증명개진후적산법시유효적。
In view of the problems of the traditional collaborative filtering recommendation algorithm based on the project of data sparseness and the low accuracy of recommendation, the thesis puts forward the HMM-ItemCF recommendation algorithm which combines Hidden Markov Model with the traditional collaborative filtering recommendation algorithm based on the project. The al-gorithm using Hidden Markov Model to all the users in the system evaluation behavior and the history of the target user behavior to carry on the overall analysis, to find the probability of the next moment a group of users with the highest score object, and the probability of occurrence of these scores with traditional objects project weighted similarity calculation method to get a new recom-mendation similarity ultimately produce results. The simulation experiment is carried out on the algorithm with an important pa-rameter in the training, and compared with other algorithms. It proves that the improved algorithm is effective.