Search

Heterogeneous cross-project defect prediction (HCPDP) aims to predict defects in new software projects using defect data from previous software projects where the source and target projects have some different metrics. Most existing methods only find linear relationships in the software defect datasets. Additionally, these methods use multiple defect datasets from different projects as source datasets. In this paper, we propose a novel method called heterogeneous cross-project defect prediction using encoder networks and transfer learning (ENTL). ENTL uses encoder networks to extract the important features from source and target datasets. Also, to minimize the negative transfer during transfer learning, we used an augmented dataset that contains pseudo-labels and the source dataset. Additionally, we have used a single dataset to train the model. To evaluate the performance of the ENTL approach, 16 datasets from four publicly available software defect projects were used. Furthermore, we compared the proposed method with four HCPDP methods namely EGW, HDP_KS, CTKCCA and EMKCA, and one WPDP method from existing literature. The proposed method on average outperforms the baseline methods in terms of PD, PF, F1-score, G-mean and AUC.

Smartphone-based approaches for Human Activity Recognition have become prevalent in recent years. Despite the amount of research undertaken in the field, issues such as cross-subject variability are still posing an obstacle to the deployment of solutions in large scale, free-living settings. Personalized methods (i.e. aiming to adapt a generic classifier to a specific target user) attempt to solve this problem. The lack of labeled data for training purposes, however, represents a major barrier. This is especially the case when taking into consideration that personalization generally requires labeled data to be user-specific. This paper presents a novel personalization method combining a semi-population based approach with user adaptation. Personalization is achieved through the following. Firstly, the proposed method identifies a subset of users from the available population as best candidates for initializing the classifier to the target user. Subsequently, a semi-population Neural Network classifier is trained using data from this subset of users. The classifier's network weights are then updated using a small amount of labeled data from the target user subsequently implementing personalization. This approach was validated on a large publicly available dataset collected in a free-living scenario. The personalized approach using the proposed method has shown to improve the overall F-score to 74.4% compared to 70.9% when using a generic non-personalized approach. Results obtained, with statistical significance being confirmed on a set of 57 users, indicate that model initialization using the semi-population approach can reduce the amount of labeled data required for personalization. As such, the proposed method for model initialization could facilitate the real-world deployment of systems implementing personalization by reducing the amount of data needed for personalization.

Research interests spanning numerous domains increasingly rely upon computational systems which can store and process a large volume of variable data that is stored at high velocity-representing a big data problem. This is particularly notable within the domain of ubiquitous and pervasive computing. This domain increasingly relies on storage and retrieval of sensor data to enable outcomes such as predictive analytics and activity recognition. Several current big data platforms exist; however, they have a range of deficiencies including lack of generic interoperability with agnostic sensors and an absence of features supporting academic research. Due to these deficiencies a custom, research oriented, high performance, and big data platform was devised and implemented. This platform is called SensorCentral and is presented within this paper. SensorCentral provides a framework which enables interoperability with a large range of agnostic sensor devices whilst simultaneously providing features which support research. Research supporting features include; facility to define experiments, ability to annotate experimental instances via purpose-built mobile applications, integrated machine learning functionality, facility to export data sets, rule-based classification and an extensible platform. The flagship implementation of this platform has been in operation for over 28 months within a University research group and has been successfully integrated with a range of sensors from a variety of manufacturers. This implementation currently stores over 850 million records and has been central to several research and industrial projects. Future work will integrate this platform into the open data initiative enabling collaboration with the international community of researchers.