International Journal of Software Engineering and Knowledge Engineering

In this paper, we address the problem of recognizing the semantic human activities through the analysis of large dataset collected from users’ sensor-based smartphones. Our approach is unique in terms of covering a large number of activities that users could possibly engage in, and considering the multi-level-based classification model. Our model has three properties that never seemed to be addressed by existing approaches dealing with the same problem. These are: (1) comprehensiveness — in terms of the activity set, (2) accuracy — in terms of the activity classification, and (3) applicability — in terms of flexibility in being applied in real-life settings. Current approaches do not tackle all these properties. When tested on realistic dataset, our multi-level-based model achieved promising results despite the large number of activities being considered. When compared to similar approaches, our approach achieved comparable results in terms of accuracy and outperformed them in terms of the activity types, environment and settings covered, comprehensiveness, and applicability.

Influence maximization is a problem that aims to select top $k$ influential nodes to maximize the spread of influence in social networks. The classical greedy-based algorithms and their improvements are relatively slow or not scalable. The efficiency of heuristic algorithms is fast but their accuracy is unacceptable. Some algorithms improve the accuracy and efficiency by consuming a large amount of memory usage. To overcome the above shortcoming, this paper proposes a fast and scalable algorithm for influence maximization, called K-paths, which utilizes the influence tree to estimate the influence spread. Additionally, extensive experiments demonstrate that the K-paths algorithm outperforms the comparison algorithms in terms of efficiency while keeping competitive accuracy.

Topic word extraction is the task of identifying single or multi-word expressions that represent the main topics of a document. In this paper, two improved algorithms for extracting and discovering topic words are proposed in the Rapid Topic word Detection (RTD) Algorithm and CategoryTextRank (CTextRank) Algorithm, which can effectively obtain information by extracting and filtering the topic words in the text. The algorithms overcome the shortcomings of traditional topic words discovering algorithms that require deep linguistic knowledge, domain or language specific annotated corpora. The two algorithms we proposed can process both short and long text. The biggest advantage of the algorithms is that they are unsupervised machine learning algorithms. They need not be trained to process text directly to get topic words. The Accuracy rate, recall rate and F-measure index have been greatly improved when using the two algorithms which show that the results obtained compare favorably with previously published results on datasets Inspec and SemEval. The first algorithm Rapid Topicword Detection improves the metrics compared to PositionRank and TextRank, the second algorithm CategoryTextRank improves the metrics compared to TextRank, SingleRank and TF-IDF.

Mutation testing aims to evaluate the fault detection capability of a test suite. This evaluation substitutes faults with mutants by transforming program code to be defective. Evidences of the relationship between the detection rates of mutants and real faults have supported the use of mutants. It has also been known that the test suite size was a significant factor affecting the relationship. Our study revealed that the selection of the mutated code was another factor affecting the relationship. We generated mutants by transforming the code modified to fix defects, while the modified code was located at three granularity levels. The experiments conducted on the defects4j dataset demonstrated that the granularity level caused a significant difference in the relationship; the detection rate of mutants was more strongly correlated with and more indicative of the fault detection capability at a fine level than at a coarse level. Moreover, the influence of the test suite size was different at each granularity level. These findings implied a strong correlation between the detection rates of mutants and real faults, independently of test suite size, when the error-prone code was located precisely.

For the test case prioritization problems, the average percent of faults detected (APFD) and its variant versions are widely used as metrics to evaluate prioritized test suite’s efficiency of fault detection. By a revisit of metrics for test case prioritization, we observe that APFD is only available for the scenarios where all test suites under evaluation contain the same number of test cases. Such a limitation is often overlooked, and lead to incorrect results when comparing fault detection efficiency of test suites with different sizes. Moreover, APFD cannot precisely illustrate the process of fault detection in the real world. Besides the APFD, most of its variants, including the NAPFD and the APFD${}_{CW}$, have similar problems. This paper points out these limitations in detail by analyzing the physical explanation of APFD series metrics formally. In order to eliminate these limitations, we propose a series of improved metrics, including the relative average percent of faults detected (RAPFD) and the relative cost-cognizant weighted average percent of faults detected (RAPFD${}_{CW}$), to evaluate the efficiency of the test suite. Furthermore, for the scenario of parallel testing, a series of metrics including the relative average percent of faults detected in parallel testing ($P$-RAPFD) and the relative cost-cognizant weighted average percent of faults detected in parallel testing ($P$-RAPFD${}_{CW}$) are proposed too. All the proposed metrics refer to both the speed of fault detection and the constraint of the testing resource. A formal analysis and some examples show that all the proposed metrics provide much more precise illustrations of the fault detection process.

Fixed Sized Candidate Set (FSCS) is the first of a series of methods proposed to enhance the effectiveness of random testing (RT) referred to as Adaptive Random Testing methods or ARTs. Since its inception, test case generation overheads have been a major drawback to the success of ART. In FSCS, the bulk of this cost is embedded in distance computations between a set of randomly generated candidate test cases and previously executed but unsuccessful test cases. Consequently, FSCS is caught in a logical trap of probing the distances between every candidate and all executed test cases before the best candidate is determined. Using data mining, however, we discovered that about 50% of all valid test cases are encountered much earlier in the distance computations process but without any benefit of a hindsight, FSCS is unable to validate them; a wild goose chase. This paper then uses this information to propose a new strategy that predictively and proactively selects valid candidates anywhere during the distance computation process without vetting every candidate. Theoretical analysis, simulations and experimental studies conducted led to a similar conclusion: 25% of the distance computations are wasteful and can be discarded without any repercussion on effectiveness.