Xianghua Fu

Artificial bee colony algorithm with gene recombination for numerical function optimization

Display Omitted An improved foraging model is designed for ABC, which can make some employed bees with high quality food source exchange information with each other.A concreted gene recombination operator (GRO) is established by recombining the different superior genes of different good individuals for generating better offspring.GRO is embedded into nine ABC methods for performance evaluation. The experimental results on 22 benchmark functions demonstrate that GRO could enhance the performance of ABC and ABC variants. Artificial bee colony (ABC) algorithm is a stochastic and population-based optimization method, which mimics the collaborative foraging behaviour of honey bees and has shown great potential to handle various kinds of optimization problems. However, ABC often suffers from slow convergence speed since its internal mechanism and solution search equation do well in exploration, but badly in exploitation. In order to solve this knotty issue, inspired by the natural phenomenon that the good individuals (solutions) always contain good genes (variables) and the effective combination of the superior genes from different good individuals could more easily produce better offspring, we introduce a novel gene recombination operator (GRO) into ABC to accelerate convergence. To be specific, in GRO, a part of good solutions in the current population are selected to produce candidate solutions by the gene combination. Especially, each good solution recombines with only one other good solution to generate only one candidate solution. In addition, GRO will be launched at the end of each generation. In order to validate its efficiency and effectiveness, GRO is embedded into nine versions of ABC, i.e., the original ABC, GABC, best-so-far ABC(BSFABC), MABC, CABC, ABCVSS, qABC, dABC and distABC, while yields GRABC, GRGABC, GRBSFABC, GRMABC, GRCABC, GRABCVSS, GRqABC, GRdABC and GRdistABC respectively. The experimental results on 22 benchmark functions demonstrate that GRO could enhance the exploitation ability of ABCs and accelerate convergence without loss of diversity.

Dynamic non-parametric joint sentiment topic mixture model

The reviews in social media are produced continuously by a large and uncontrolled number of users. To capture the mixture of sentiment and topics simultaneously in reviews is still a challenging task. In this paper, we present a novel probabilistic model framework based on the non-parametric hierarchical Dirichlet process (HDP) topic model, called non-parametric joint sentiment topic mixture model (NJST), which adds a sentiment level to the HDP topic model and detects sentiment and topics simultaneously from reviews. Then considered the dynamic nature of social media data, we propose dynamic NJST (dNJST) which adds time decay dependencies of historical epochs to the current epochs. Compared with the existing sentiment topic mixture models which are based on latent Dirichlet allocation (LDA), the biggest difference of NJST and dNJST is that they can determine topic number automatically. We implement NJST and dNJST with online variational inference algorithms, and incorporate the sentiment priors of words into NJST and dNJST with HowNet lexicon. The experiment results in some Chinese social media dataset show that dNJST can effectively detect and track dynamic sentiment and topics.

Learning distributed word representation with multi-contextual mixed embedding

Learning distributed word representations has been a popular method for various natural language processing applications such as word analogy and similarity, document classification and sentiment analysis. However, most existing word embedding models only exploit a shallow slide window as the context to predict the target word. Because the semantic of each word is also influenced by its global context, as the distributional models usually induced the word representations from the global co-occurrence matrix, the window-based models are insufficient to capture semantic knowledge. In this paper, we propose a novel hybrid model called mixed word embedding (MWE) based on the well-known word2vec toolbox. Specifically, the proposed MWE model combines the two variants of word2vec, i.e., SKIP-GRAM and CBOW, in a seamless way via sharing a common encoding structure, which is able to capture the syntax information of words more accurately. Furthermore, it incorporates a global text vector into the CBOW variant so as to capture more semantic information. Our MWE preserves the same time complexity as the SKIP-GRAM. To evaluate our MWE model efficiently and adaptively, we study our model on linguistic and application perspectives with both English and Chinese dataset. For linguistics, we conduct empirical studies on word analogies and similarities. The learned latent representations on both document classification and sentiment analysis are considered for application point of view of this work. The experimental results show that our MWE model is very competitive in all tasks as compared with the state-of-the-art word embedding models such as CBOW, SKIP-GRAM, and GloVe.

Dynamic Online HDP model for discovering evolutionary topics from Chinese social texts

User-generated content such as online reviews in social media evolve rapidly over time. To better understand the social media content, users not only want to examine what the topics are, but also want to discover the topic evolution patterns. In this paper, we propose a Dynamic Online Hierarchical Dirichlet Process model (DOHDP) to discover the evolutionary topics for Chinese social texts. In our DOHDP model, the evolutionary processes of topics are considered as evolutions in two levels, i.e. inter-epoch level and intra-epoch level. In inter-epoch level, the corpus of each epoch is modeled with an online HDP topic model, and the social texts are generated in a sequence mode. In the intra-epoch level, the time dependencies of historical epochs are modeled with an exponential decay function in which more recent epochs have a relatively stronger influence on the model parameters than the earlier epoch. Furthermore, we implement our DOHDP model using a two-phase online variational algorithm. Through comparing our DOHDP model with other related topic models on Chinese social media dataset Tianya-80299, the experiment results show that DOHDP model provides the best performance for discovering the evolutionary topics of Chinese social texts.

Combine HowNet lexicon to train phrase recursive autoencoder for sentence-level sentiment analysis

Abstract Detecting sentiment of sentences in online reviews is still a challenging task. Traditional machine learning methods often use bag-of-words representations which cannot properly capture complex linguistic phenomena in sentiment analysis. Recently, recursive autoencoder (RAE) methods have been proposed for sentence-level sentiment analysis. They use word embedding to represent each word, and learn compositional vector representation of phrases and sentences with recursive autoencoders. Although RAE methods outperform other state-of-the-art sentiment prediction approaches on commonly used datasets, they tend to generate very deep parse trees, and need a large amount of labeled data for each node during the process of learning compositional vector representations. Furthermore, RAE methods mainly combine adjacent words in sequence with a greedy strategy, which make capturing semantic relations between distant words difficult. To solve these issues, we propose a semi-supervised method which combines HowNet lexicon to train phrase recursive autoencoders (we call it CHL-PRAE). CHL-PRAE constructs the phrase recursive autoencoder (PRAE) model at first. Then the model calculates the sentiment orientation of each node with the HowNet lexicon, which acts as sentiment labels, when we train the softmax classifier of PRAE. Furthermore, our CHL-PRAE model conducts bidirectional training to capture global information. Compared with RAE and some supervised methods such as support vector machine (SVM) and naive Bayesian on English and Chinese datasets, the experiment results show that CHL-PRAE can provide the best performance for sentence-level sentiment analysis.

Exploring A Trust Based Recommendation Approach for Videos in Online Social Network

With the rapid development of social network, more and more users watch videos through social network, such as Sina Weibo. Traditional video recommendation algorithms aim at online video systems and they neglect the social relationship and propagation features in social network. The interaction information among users in social network could help improve the effect of video recommendation in social network. This paper mainly focuses on the problem that current video recommendation methods for videos in online social network can not meet the needs of the users. To address this challenge, we propose a new trust based video recommendation approach including a user discovery model and a video discovery model in this paper. To discover influential users of the target user, we divide the other users into direct influential users and indirect influential users. We compute the trust between the target user and each of his/her influential users based on user similarity, friendship and interaction. In the video discovery model, we calculate the video trust based on the video rating and video activity. Through combing the user discovery model and video discovery model, we present our trust based recommendation algorithm for videos in social network. The experimental results demonstrate that our approach can outperform two classical video recommendation algorithms, in terms of precision, recall and F1-measure.

Modified Gbest-guided artificial bee colony algorithm with new probability model

Artificial bee colony (ABC) is a very effective and efficient swarm-based intelligence optimization algorithm, which simulates the collective foraging behavior of the honey bees. However, ABC has strong exploration ability but poor exploitation ability because its solution search equation performs well in exploration but badly in exploitation. In order to enhance the exploitation ability and obtain a better balance between exploitation and exploration, in this paper, a novel search strategy which exploits the valuable information of the current best solution and a novel probability model which makes full use of the other good solutions on onlooker bee phase are proposed. To be specific, in the novel search strategy, a parameter P is used to control which search equation to be used, the original search equation of ABC or the new proposed search equation. The new proposed search equation utilizes the useful information from the current best solution. In the novel probability model, the selected probability of the good solution is absolutely significantly larger than that of the bad solution, which makes sure the good solutions can attract more onlooker bees to search. We put forward a new ABC variant, named MPGABC by combining the novel search strategy and probability model with the basic framework of ABC. Through the comparison of MPGABC and some other state-of-the-art ABC variants on 22 benchmark functions, 22 CEC2011 real-world optimization problems and 28 CEC2013 real-parameter optimization problems, the experimental results show that MPGABC is better than or at least comparable to the competitors on most of benchmark functions and real-world problems.

A novel multi-objective evolutionary algorithm for recommendation systems

Nowadays, the recommendation algorithm has been used in lots of information systems and Internet applications. The recommendation algorithm can pick out the information that users are interested in. However, most traditional recommendation algorithms only consider the precision as the evaluation metric of the performance. Actually, the metrics of diversity and novelty are also very important for recommendation. Unfortunately, there is a conflict between precision and diversity in most cases. To balance these two metrics, some multi-objective evolutionary algorithms are applied to the recommendation algorithm. In this paper, we firstly put forward a kind of topic diversity metric. Then, we propose a novel multi-objective evolutionary algorithm for recommendation systems, called PMOEA. In PMOEA, we present a new probabilistic genetic operator. Through the extensive experiments, the results demonstrate that the combination of PMOEA and the recommendation algorithm can achieve a good balance between precision and diversity. A new topic diversity indicator is introduced, which can be used to measure various kinds of items in a recommendation list.A new probabilistic multi-objective evolutionary algorithm (PMOEA) is presented, which is suitable for the recommendation systems.A new crossover operator is proposed to generate new solution, called the multi-parent probability genetic operator.The experimental results show that PMOEA can achieve a good balance between precision and diversity.

Accurate RFID localization algorithm with particle swarm optimization based on reference tags

RFID technology has been widely used for object tracking in indoor environment due to their low cost and convenience for deployment. Existing RFID localization approaches rely on signal strength to measure the distance between RFID reader and tags. However, because of the environmental complexity and inferences, the measurement of distance by signal strength is not accurate, which causes large error in localization. In this paper we develop a novel algorithm to improve the RFID localization accuracy. Our algorithm is based on particle swarm optimization. More importantly, we add reference tags in the deployment, and design a parameter named Correction Factor in PSO to measure the distances more accurately by signal strength. The result shows that compared with the previous method without the correction factor, our proposed approach can increase the accuracy by 50%. This method has good application prospect in equipment management.

A video recommendation algorithm based on the combination of video content and social network

Recently, social network has been one of the biggest information exchange platforms of the Internet. Moreover, the users in social network used to watch videos through social network application. To provide a proper recommended video list, the video recommendation algorithm for social network is becoming a hot research issue. On one hand, more and more researchers introduce the concept of trust into video recommendation algorithms. However, most of them only select the trust friends based on the similarity and neglect the characteristics of social network. On the other hand, most previous video recommendation algorithms are only based on the number that a video is viewed to evaluate a videos quality. They do not make good use of the social relationship in social network and the videos reputation. This paper mainly focuses on the challenge that the effectiveness and performance of current video recommendation algorithm in social network cannot satisfy the users. In this paper, we propose a novel video recommendation algorithm based on the combination of video content and social network. Our proposed algorithm consists of the trust friends computing model and videos quality evaluation model. The trust friends computing method takes into account similarity between users, interaction between users, and the active degree of a user. In our videos quality evaluation model, we combine the acceptance ratio of a video with a videos reputation. The video can be given an appropriate rating score through this model. We design corresponding trust friends computing algorithm and video recommendation algorithm respectively for two proposed models. Our integral video recommendation algorithm consists of these two algorithms. The experimental results indicate that the performance and effectiveness of our algorithm are better than those of two classical video recommendation algorithms (i.e., user‐based collaborative filtering algorithm and TBR‐d algorithm), in terms of precision, recall and F1‐measure. Copyright