In software engineering, defect prediction is significantly important and challenging. The main task is to predict the defect proneness of the modules. It helps developers find bugs effectively and prioritize their testing efforts. At present, a lot of valuable researches have been done on this topic. However, few studies take into account the impact of time factors on the prediction results. Therefore, in this paper, we propose an improved Elman neural network model to enhance the adaptability of the defect prediction model to the time-varying characteristics. Specifically, we optimized the initial weights and thresholds of the Elman neural network by incorporating adaptive step size in the Cuckoo Search (CS) algorithm. We evaluated the proposed model on 7 projects collected from public PROMISE repositories. The results suggest that the contribution of the improved CS algorithm to Elman neural network model is prominent, and the prediction performance of our method is better than that of 5 baselines in terms of F-measure and Cliff’s Delta values. The F-measure values are generally increased with a maximum growth rate of 49.5% for the POI project.

Quality of Service (QoS) prediction for Web services is a hot research topic in the field of services computing. Recently, representation learning of heterogeneous networks has attracted much attention, and specifically the relationship between users and services, as a typical heterogeneous network in which heterogeneity and rich semantic information provide a new perspective for QoS prediction. This paper proposes a novel QoS Prediction scheme based on a heterogeneous graph attention network. Our method first unitizes the user's location information to construct an attributed user-service network. Then, considering the difference between nodes and links in the latter network, we model a heterogeneous graph neural network based on a hierarchical attention machine (HGN2HIA) that includes node-and semanticlevel attentions. Specifically, node-level attention aims to learn the importance between a node and its meta-path-based neighbors, while semantic-level attention learns the importance of different meta-paths. Finally, user embedding will be generated by aggregating features from meta-path-based neighbors in a hierarchical manner, used for QoS prediction. Experimental results on the public WS-Dream dataset demonstrate the superior performance of the proposed model over the current state-of-the-art methods, with NMAE and RMSE metrics reduced by at least 2.56% and 1.3%, respectively. Furthermore, the experimental results highlight that node-level attention contributes more than semantic-level. Overall, we demonstrate that introducing these attention levels improves the QoS prediction performance.

Cross-project defect prediction (CPDP) is a mainstream method estimating the most defect-prone components of software with limited historical data. Several studies investigate how software metrics are used and how modeling techniques influence prediction performance. However, the software’s metrics diversity impact on the predictor remains unclear. Thus, this paper aims to assess the impact of various metric sets on CPDP and investigate the feasibility of CPDP with hybrid metrics. Based on four software metrics types, we investigate the impact of various metric sets on CPDP in terms of F-measure and statistical methods. Then, we validate the dominant performance of CPDP with hybrid metrics. Finally, we further verify the CPDP-OSS feasibility built with three types of metrics (orient-object, semantic, and structural metrics) and challenge them against two current models. The experimental results suggest that the impact of different metric sets on the performance of CPDP is significantly distinct, with semantic and structural metrics performing better. Additionally, trials indicate that it is helpful for CPDP to increase the software’s metrics diversity appropriately, as the CPDP-OSS improvement is up to 53.8%. Finally, compared with two baseline methods, TCA+ and TDSelector, the optimized CPDP model is viable in practice, and the improvement rate is up to 50.6% and 25.7%, respectively.

With the increasing proliferation of malicious code, the camouflage of malicious code is more difficult to cope with. Traditional malicious code detection techniques based on byte comparison have limited accuracy. Detection techniques based on traditional machine learning are highly dependent on feature selection, and the quality of the classifier directly affects the detection results; this increases the difficulty of accurately distinguishing the types of malicious code. To address these problems, a deep neural network-based malicious code detection method is proposed in this work. First, the code binary file is transformed into a corresponding gray-scale image, and then the enhanced RGBA image is formed by using an image enhancement scheme based on information entropy and code file structure. Afterwards, a convolutional neural network is used. The network extracts high-dimensional features of the enhanced code image, detects the malicious code, and classifies the malicious code. The experimental results show that the proposed method distinguishes malicious code with 98.83% detection accuracy. Its classification accuracy is 97.74% (with positive samples) and 98.85% (without positive samples). These high levels of accuracy are suitable for current complex and changeable malicious code environments, and can provide a new solution for the current malicious code detection field.