Most existing visual sentiment analysis methods mainly construct visual sentiment feature representation based on the whole image. However, the local regions with objects in the image are able to highlight the sentiment better. Concerning the problem of ignorance of local regions sentiment representation in visual sentiment analysis, a visual sentiment analysis method by combining global and local regions of image was proposed. Image sentiment representation was mined by combining a whole image with local regions of the image. Firstly, an object detection model was used to locate the local regions with objects in the image. Secondly, the sentiment features of the local regions with objects were extracted by deep neural network. Finally, the deep features extracted from the whole image and the local region features were utilized to jointly train the image sentiment classifier and predict the sentiment polarity of the image. Experimental results show that the classification accuracy of the proposed method reaches 75.81% and 78.90% respectively on the real datasets TwitterⅠand TwitterⅡ, which is higher than the accuracy of sentiment analysis methods based on features extracted from the whole image or features extracted from the local regions of image.

Concerning the problem that clustering results are usually required to be balanced in practical applications, a Balanced Clustering algorithm based on Simulated annealing and Greedy strategy (BCSG) was proposed. The algorithm includes two steps:Simulated Annealing Clustering Initialization (SACI) and Balanced Clustering based on Greedy Strategy (BCGS) to improve clustering effectiveness with less time cost. First of all, K suitable data points of data set were located based on simulated annealing as the initial point of balanced clustering, and the nearest data points to each center point were added into the cluster where it belongs in stages greedily until the cluster size reach the upper limit. A series of experiments carried on six UCI real datasets and two public image datasets show that the balance degree can be increased by more than 50 percentage points compared with Fuzzy C-Means when the number of clusters is large, and the accuracy of clustering result is increased by 8 percentage points compared with Balanced K-Means and BCLS (Balanced Clustering with Least Square regression) which have good balanced clustering performance. Meanwhile, the time complexity of the BCSG is also lower, the running time is decreased by nearly 40 percentage points on large datasets compared with Balanced K-Means. BCSG has better clustering effectiveness with less time cost than other balanced clustering algorithms.

In recent years, more and more multimedia contents were used on social media to share users' experiences and emotions. Compared to single text or image, the complementation of text and image can further fully reveal the real emotion of users. Concerning the sentiment shortage of single text or image, a method based on Convolutional Neural Network (CNN) was proposed for multimedia sentiment analysis. In order to explore the influence of semantic representation in different level, image features were combined with different level (word-level, phrase-level and sentence-level) text features to construct CNN. The experimental results on two real-world datasets demonstrate that the proposed method gets more accurate prediction on multimedia sentiment analysis by capturing the internal relations between text and image.

In user-based collaborative filtering recommendation based on social relations, sometimes the ratings for the target items can not be predicted. Whats more, in traditional item-based collaborative filtering, there are still some items which are not in the same class with the target item and not suitable to be references for predicting ratings. To handle these problems, two new algorithms of collaborative filtering recommendation were proposed, in which the tags of scenic spots type were introduced to compute the similarity between two scenic spots. The experimental results on the data set of scenic spots ratings show that, compared with the user-based collaborative filtering recommendation algorithms based on social relations, the algorithm based on the social relation and tag can increase the accuracy and the coverage by 10% and 4% respectively, and compared with the item-based collaborative filtering recommendation algorithms, the collaborative filtering recommendation algorithm based on item and tag can increase the accuracy by 15%, it also shows that introducing the tags of scenic spots type can make the computation of the similarity between two scenic spots more accurate.