Collaborative Filtering Service Recommendation Based on a Novel Similarity Computation Method
Recently, collaborative filtering-based methods are widely used for service recommendation. QoS attribute value based collaborative filtering service recommendation includes two important steps. One is the similarity computation, and the other is the prediction for the QoS attribute value, which the user has not experienced. In some previous studies, the similarity computation methods and prediction methods are not accurate. The performances of some methods need to be improved. In this paper, we propose a ratio-based method to calculate the similarity. We can get the similarity between users or between items by comparing the attribute values directly. Based on our similarity computation method, we propose a new method to predict the unknown value. By comparing the values of a similar service and the current service that are invoked by common users, we can obtain the final prediction result. The performance of the proposed method is evaluated through a large data set of real web services. Experimental results show that our method obtains better prediction precision, lower mean absolute error (MAE) and faster computation time than various reference schemes considered.
- Presently, the Pearson correlation coefficient (PCC) and cosine (COS) methods are commonly applied to calculate the similarity.
DISADVANTAGES OF EXISTING SYSTEM:
- Pearson correlation coefficient (PCC) and cosine (COS) methods have limited accuracy.
- PCC method does not take the differences of QoS attributes values given by different users into account. Although the COS method can measure the angles of the vectors, which are composed by the users or services, it neglects the lengths of the vectors.
- In this paper, we propose a new method to calculate the similarity.
- Generally, the QoS attributes experienced by the user are given in the form of numerical values, and these values are non-negative. The similarity represents the degree of two objects’ consistency. We can use the ratio of two values to express the consistency.
- The ratio of two attribute values which is the results of two users invoking the same item reflects the users’ consistency on this item, i.e., the single similarity. Summing up all the single similarities together and getting the average, we can obtain the final similarity between two users.
ADVANTAGES OF PROPOSED SYSTEM:
- Our method is applicable to all kinds of QoS attributes which are given in the numerical values. However, some of the qualitative and subjective QoS attributes are expressed in non-numerical value, such as “very good”, “good”, and so on. According to certain rules, these evaluations can be transformed into numerical values, and then our method can be used.
- The recommendation system can recommend appropriate service(s) to the user according to given conditions. Here, the specific condition given by a user may be constrained by multiple objectives
- Save a lot of time and energy
- System : Pentium Dual Core.
- Hard Disk : 120 GB.
- Monitor : 15’’ LED
- Input Devices : Keyboard, Mouse
- Ram : 1 GB
- Operating system : Windows 7.
- Coding Language : JAVA/J2EE
- Tool : Netbeans 7.2.1
- Database : MYSQL
Xiaokun Wu, Bo Cheng, and Junliang Chen, “Collaborative Filtering Service Recommendation Based on a Novel Similarity Computation Method”, IEEE TRANSACTIONS ON SERVICE COMPUTING, VOL.10, NO.3, May-June 2017.