Chunping Li

Workflow simulation for operational decision support using event graph through process mining

It is increasingly common to see computer-based simulation being used as a vehicle to model and analyze business processes in relation to process management and improvement. While there are a number of business process management (BPM) and business process simulation (BPS) methodologies, approaches and tools available, it is more desirable to have a systemic BPS approach for operational decision support, from constructing process models based on historical data to simulating processes for typical and common problems. In this paper, we have proposed a generic approach of BPS for operational decision support which includes business processes modeling and workflow simulation with the models generated. Processes are modeled with event graphs through process mining from workflow logs that have integrated comprehensive information about the control-flow, data and resource aspects of a business process. A case study of a credit card application is presented to illustrate the steps involved in constructing an event graph. The evaluation detail is also given in terms of precision, generalization and robustness. Based on the event graph model constructed, we simulate the process under different scenarios and analyze the simulation logs for three generic problems in the case study: 1) suitable resource allocation plan for different case arrival rates; 2) teamwork performance under different case arrival rates; and 3) evaluation and prediction for personal performances. Our experimental results show that the proposed approach is able to model business processes using event graphs and simulate the processes for common operational decision support which collectively play an important role in process management and improvement.

Combining Lexical and Semantic Features for Short Text Classification

Abstract In this paper, we propose a novel approach to classify short texts by combining both their lexical and semantic features. We present an improved measurement method for lexical feature selection and furthermore obtain the semantic features with the background knowledge repository which covers target category domains. The combination of lexical and semantic features is achieved by mapping words to topics with different weights. In this way, the dimensionality of feature space is reduced to the number of topics. We here use Wikipedia as background knowledge and employ Support Vector Machine (SVM) as classifier. The experiment results show that our approach has better effectiveness compared with existing methods for classifying short texts.

A topical PageRank based algorithm for recommender systems

In this paper, we propose a Topical PageRank based algorithm for recommender systems, which aim to rank products by analyzing previous user-item relationships, and recommend top-rank items to potentially interested users. We evaluate our algorithm on MovieLens dataset and empirical experiments demonstrate that it outperforms other state-of-the-art recommending algorithms.

Social recommendation incorporating topic mining and social trust analysis

We study the problem of social recommendation incorporating topic mining and social trust analysis. Different from other works related to social recommendation, we merge topic mining and social trust analysis techniques into recommender systems for finding topics from the tags of the items and estimating the topic-specific social trust. We propose a probabilistic matrix factorization (TTMF) algorithm and try to enhance the recommendation accuracy by utilizing the estimated topic-specific social trust relations. Moreover, TTMF is also convenient to solve the item cold start problem by inferring the feature (topic) of new items from their tags. Experiments are conducted on three different data sets. The results validate the effectiveness of our method for improving recommendation performance and its applicability to solve the cold start problem.

Stock temporal prediction based on time series motifs

Recent researches pay more attention to stock tendency prediction, which various machine learning approaches have been proposed. In this paper, we propose an algorithm to discover self-correlation of stock price in virtue of the notion of time series motifs, by viewing stock price sequences as time series. Generally, time series motif is a pattern appearing frequently in a time sequence, useful to forecast the stock temporal tendencies and prices as a reliable part in time series. In the proposed approach, we firstly search for one part of time series motifs using ordinal comparison and k-NN clustering algorithm, and then attempt to discover the correlation between motifs and subsequences connected behind them. Experimental results demonstrate the positive contribution of time series motifs, the acceptable prediction accuracy, and priority of our algorithm.

Augmenting Chinese Online Video Recommendations by Using Virtual Ratings Predicted by Review Sentiment Classification

In this paper we aim to resolve the recommendation problem by using the virtual ratings in online environments when user rating information is not available. As a matter of fact, in most of current websites especially the Chinese video-sharing ones, the traditional pure rating based collaborative filtering recommender methods are not fully qualified due to the sparsity of rating data. Motivated by our prior work on the investigation of user reviews that broadly appear in such sites, we hence propose a new recommender algorithm by fusing a self-supervised emoticon-integrated sentiment classification approach, by which the missing User-Item Rating Matrix can be substituted by the virtual ratings which are predicted by decomposing user reviews as given to the items. To test the algorithm’s practical value, we have first identified the self-supervised sentiment classification’s higher performance by comparing it with a supervised approach. Moreover, we conducted a statistic evaluation method to show the effectiveness of our recommender system on improving Chinese online video recommendations’ accuracy.

Big data analytics for security and criminal investigations

Applications of various data analytics technologies to security and criminal investigation during the past three decades have demonstrated the inception, growth, and maturation of criminal analytics. We first identify five cutting‐edge data mining technologies such as link analysis, intelligent agents, text mining, neural networks, and machine learning. Then, we explore their recent applications to the criminal analytics domain, and discuss the challenges arising from these innovative applications. We also extend our study to big data analytics which provides some state‐of‐the‐art technologies to reshape criminal investigations. In this paper, we review the recent literature, and examine the potentials of big data analytics for security intelligence under a criminal analytics framework. We examine some common data sources, analytics methods, and applications related to two important aspects of social network analysis namely, structural analysis and positional analysis that lay the foundation of criminal analytics. Another contribution of this paper is that we also advocate a novel criminal analytics methodology that is underpinned by big data analytics. We discuss the merits and challenges of applying big data analytics to the criminal analytics domain. Finally, we highlight the future research directions of big data analytics enhanced criminal investigations. WIREs Data Mining Knowl Discov 2017, 7:e1208. doi: 10.1002/widm.1208

A random-walk based recommendation algorithm considering item categories

Abstract Recommender systems aim at recommending information items or social elements that are likely to be of interest to users. In this paper, we propose a recommendation algorithm which takes into account users preference on item categories, and computes rank scores in different categories for each item, in order to make suggestions based on both users previous interactions and item contents. By considering item categories and user preference, we are able to avoid the dominance of some popular items. Empirical experiments on MovieLens dataset demonstrate that the algorithm outperforms other state-of-the-art recommendation algorithms.

Cross-Domain Knowledge Transfer Using Semi-supervised Classification

Traditional text classification algorithms are based on a basic assumption: the training and test data should hold the same distribution. However, this identical distribution assumption is always violated in real applications. Due to the distribution of test data from target domain and the distribution of training data from auxiliary domain are different, we call this classification problem cross-domain classification. Although most of the training data are drawn from auxiliary domain, we still can obtain a few training data drawn from target domain. To solve the cross-domain classification problem in this situation, we propose a two-stage algorithm which is based on semi-supervised classification. We firstly utilizes labeled data in target domain to filter the support vectors of the auxiliary domain, then uses filtered data and labeled data from target domain to construct a classifier for the target domain. The experimental evaluation on real-world text classification problems demonstrates encouraging results and validates our approach.

Turn Waste into Wealth: On Simultaneous Clustering and Cleaning over Dirty Data

Dirty data commonly exist. Simply discarding a large number of inaccurate points (as noises) could greatly affect clustering results. We argue that dirty data can be repaired and utilized as strong supports in clustering. To this end, we study a novel problem of clustering and repairing over dirty data at the same time. Referring to the minimum change principle in data repairing, the objective is to find a minimum modification of inaccurate points such that the large amount of dirty data can enhance the clustering. We show that the problem can be formulated as an integer linear programming (ILP) problem. Efficient approximation is then devised by a linear programming (LP) relaxation. In particular, we illustrate that an optimal solution of the LP problem can be directly obtained without calling a solver. A quadratic time approximation algorithm is developed based on the aforesaid LP solution. We further advance the algorithm to linear time cost, where a trade-off between effectiveness and efficiency is enabled. Empirical results demonstrate that both the clustering and cleaning accuracies can be improved by our approach of repairing and utilizing the dirty data in clustering.