Research On Service Recommendation Method Based On Context Awareness

With the rapid development of economical and social construction and information technology,how to filter user interest information from massive data becomes one of the challenging issues in the "big data" era.The recommendation system came into being,it based on specific algorithms and indicators,mining and analyzing the user's historical data,and recommending for users from the massive data to meet their interests and needs of information actively.However,traditional recommendation algorithms exist some problems,such as do not consider user context information,cold start,and data sparsity,which reduces the accuracy and effectiveness of the recommendation.This paper takes user context information into account in the recommendation algorithm,modeling user interest,and recommending similar users after clustering.Compared with the traditional recommendation algorithm,it can effectively improve the user ' s recommendation satisfaction and accuracy.It enables users with different needs to receive more personalized service recommendations,which is of great significance in practical applications and research projects.The main work and contributions of this paper are summarized as follows:(1)Considering the effect of context information on recommendation,a context clustering collaborative filtering based on K-means(KCCF)recommendation proposed.In response to the problems of do not consider user context information and data sparsity existing in traditional collaborative filtering algorithms,comprehensive consideration is given to the user's situation,combining the advantages of context awareness and collaborative filtering algorithms,user and project context information are used as parameters to construct a user-item-situation scoring matrix.The K-means algorithm is used to cluster users with common interests,then the user-based collaborative filtering recommendation algorithm is applied to the clusters generated by the clustering algorithm to achieve the rating prediction and Top-N recommendation.(2)Considering the influence of time factors on the recommendation effect,the above algorithm is further improved and a context clustering collaborative filtering based on K-means and time-weighted(KTCF)recommendation method is proposed.Based on the above algorithm,the time factor is added when the user predicts the score and the similarity calculation,and the score weights and the similarity matrix are re-calculated and distributed,and achieved the Top-N recommendation.(3)As a comparison algorithm,an improved collaborative filtering recommendation based on context awareness and neural network(CNCF)is analyzed.The neural network is used to mine the hidden features in the context information,and the name of the data set is used to perform feature mining using the convolutional neural network and other contexts.The information is built into the embedded matrix.After training,the user association matrix and the project association matrix are obtained respectively.After the scoring is optimized and the loss optimization MAE is performed,the Top-N recommendation is implemented.(4)The algorithm design and experiment of the above three context-aware service recommendation methods and traditional collaborative filtering recommendation methods were performed.Using MovieLens data set as data source.During the experiment process,different values were taken for similar neighbors and recommended numbers.Recall,Precision,F1,MAE,and other indicators were used to evaluate the algorithm in both score prediction and Top-N recommendation.The experimental results show that the context clustering collaborative filtering based on K-means and the context clustering collaborative filtering based on K-means and time-weighted and the improved collaborative filtering recommendation based on context awareness and neural network that proposed in this paper have higher score prediction accuracy and higher recommendation efficiency.