Kok-Leong Ong

Enhancing the Effectiveness of Clustering with Spectra Analysis

For many clustering algorithms, such as K-Means, EM, and CLOPE, there is usually a requirement to set some parameters. Often, these parameters directly or indirectly control the number of clusters, that is, k, to return. In the presence of different data characteristics and analysis contexts, it is often difficult for the user to estimate the number of clusters in the data set. This is especially true in text collections such as Web documents, images, or biological data. In an effort to improve the effectiveness of clustering, we seek the answer to a fundamental question: How can we effectively estimate the number of clusters in a given data set? We propose an efficient method based on spectra analysis of eigenvalues (not eigenvectors) of the data set as the solution to the above. We first present the relationship between a data set and its underlying spectra with theoretical and experimental results. We then show how our method is capable of suggesting a range of k that is well suited to different analysis contexts. Finally, we conclude with further empirical results to show how the answer to this fundamental question enhances the clustering process for large text collections.

Online mining of frequent sets in data streams with error guarantee

For most data stream applications, the volume of data is too huge to be stored in permanent devices or to be thoroughly scanned more than once. It is hence recognized that approximate answers are usually sufficient, where a good approximation obtained in a timely manner is often better than the exact answer that is delayed beyond the window of opportunity. Unfortunately, this is not the case for mining frequent patterns over data streams where algorithms capable of online processing data streams do not conform strictly to a precise error guarantee. Since the quality of approximate answers is as important as their timely delivery, it is necessary to design algorithms to meet both criteria at the same time. In this paper, we propose an algorithm that allows online processing of streaming data and yet guaranteeing the support error of frequent patterns strictly within a user-specified threshold. Our theoretical and experimental studies show that our algorithm is an effective and reliable method for finding frequent sets in data stream environments when both constraints need to be satisfied.

Training loads and injury risk in Australian football—differing acute: chronic workload ratios influence match injury risk

Aims (1) To investigate whether a daily acute:chronic workload ratio informs injury risk in Australian football players; (2) to identify which combination of workload variable, acute and chronic time window best explains injury likelihood. Methods Workload and injury data were collected from 53 athletes over 2 seasons in a professional Australian football club. Acute:chronic workload ratios were calculated daily for each athlete, and modelled against non-contact injury likelihood using a quadratic relationship. 6 workload variables, 8 acute time windows (2–9 days) and 7 chronic time windows (14–35 days) were considered (336 combinations). Each parameter combination was compared for injury likelihood fit (using R2). Results The ratio of moderate speed running workload (18–24 km/h) in the previous 3 days (acute time window) compared with the previous 21 days (chronic time window) best explained the injury likelihood in matches (R2=0.79) and in the immediate 2 or 5 days following matches (R2=0.76–0.82). The 3:21 acute:chronic workload ratio discriminated between high-risk and low-risk athletes (relative risk=1.98–2.43). Using the previous 6 days to calculate the acute workload time window yielded similar results. The choice of acute time window significantly influenced model performance and appeared to reflect the competition and training schedule. Conclusions Daily workload ratios can inform injury risk in Australian football. Clinicians and conditioning coaches should consider the sport-specific schedule of competition and training when choosing acute and chronic time windows. For Australian football, the ratio of moderate speed running in a 3-day or 6-day acute time window and a 21-day chronic time window best explained injury risk.

A Comparison of Recruitment Methods for an mHealth Intervention Targeting Mothers: Lessons from the Growing Healthy Program

Background Mobile health (mHealth) programs hold great promise for increasing the reach of public health interventions. However, mHealth is a relatively new field of research, presenting unique challenges for researchers. A key challenge is understanding the relative effectiveness and cost of various methods of recruitment to mHealth programs. Objective The objectives of this study were to (1) compare the effectiveness of various methods of recruitment to an mHealth intervention targeting healthy infant feeding practices, and (2) explore factors influencing practitioner referral to the intervention. Methods The Growing healthy study used a quasi-experimental design with an mHealth intervention group and a concurrent nonrandomized comparison group. Eligibility criteria included: expectant parents (>30 weeks of gestation) or parents with an infant <3 months old, ability to read and understand English, own a mobile phone, ≥18 years old, and living in Australia. Recruitment to the mHealth program consisted of: (1) practitioner-led recruitment through Maternal and Child Health nurses, midwives, and nurses in general practice; (2) face-to-face recruitment by researchers; and (3) online recruitment. Participants’ baseline surveys provided information regarding how participants heard about the study, and their sociodemographic details. Costs per participant recruited were calculated by taking into account direct advertising costs and researcher time/travel costs. Practitioner feedback relating to the recruitment process was obtained through a follow-up survey and qualitative interviews. Results A total of 300 participants were recruited to the mHealth intervention. The cost per participant recruited was lowest for online recruitment (AUD

Prediction of wool knitwear pilling propensity using support vector machines

14) and highest for practice nurse recruitment (AUD

SCLOPE: An Algorithm for Clustering Data Streams of Categorical Attributes

586). Just over half of the intervention group (50.3%, 151/300) were recruited online over a 22-week period compared to practitioner recruitment (29.3%, 88/300 over 46 weeks) and face-to-face recruitment by researchers (7.3%, 22/300 over 18 weeks). No significant differences were observed in participant sociodemographic characteristics between recruitment methods, with the exception that practitioner/face-to-face recruitment resulted in a higher proportion of first-time parents (68% versus 48%, P=.002). Less than half of the practitioners surveyed reported referring to the program often or most of the time. Key barriers to practitioner referral included lack of time, difficulty remembering to refer, staff changes, lack of parental engagement, and practitioner difficulty in accessing the app. Conclusions Online recruitment using parenting-related Facebook pages was the most cost effective and timely method of recruitment to an mHealth intervention targeting parents of young infants. Consideration needs to be given to addressing practitioner barriers to referral, to further explore if this can be a viable method of recruitment.

Non-repudiation in an agent-based electronic commerce system

The propensity of wool knitwear to form entangled fiber balls, known as pills, on the surface is affected by a large number of factors. This study examines, for the first time, the application of the support vector machine (SVM) data mining tool to the pilling propensity prediction of wool knitwear. The results indicate that by using the binary classification method and the radial basis function (RBF) kernel function, the SVM is able to give high pilling propensity prediction accuracy for wool knitwear without data over-fitting. The study also found that the number of records available for each pill rating greatly affects the learning and prediction capability of SVM models.

Assessing User Engagement of an mHealth Intervention: Development and Implementation of the Growing Healthy App Engagement Index

Clustering is a difficult problem especially when we consider the task in the context of a data stream of categorical attributes. In this paper, we propose SCLOPE, a novel algorithm based on CLOPE’s intuitive observation about cluster histograms. Unlike CLOPE however, our algo- rithm is very fast and operates within the constraints of a data stream environment. In particular, we designed SCLOPE according to the recent CluStream framework. Our evaluation of SCLOPE shows very promising results. It consistently outperforms CLOPE in speed and scalability tests on our data sets while maintaining high cluster purity; it also supports cluster analysis that other algorithms in its class do not.

2Loud ?: Community mapping of exposure to traffic noise with mobile phones

ABECOS is an agent-based e-commerce system under development at the Nanyang Technological University. A key factor in making this system usable in practice is strict security control. One aspect of security is the provision of non-repudiation services. As protocols for non-repudiation have focused on message non-repudiation, its adaptation to afford non-repudiation in a communication session for two agents in ABECOS is inefficient. In this work, we investigate and propose a protocol for enforcing non-repudiation in a session. The protocol is believed to be applicable in any e-commerce system; agent- or not agent-based.

Preventing obesity in infants: the Growing healthy feasibility trial protocol

Background Childhood obesity is an ongoing problem in developed countries that needs targeted prevention in the youngest age groups. Children in socioeconomically disadvantaged families are most at risk. Mobile health (mHealth) interventions offer a potential route to target these families because of its relatively low cost and high reach. The Growing healthy program was developed to provide evidence-based information on infant feeding from birth to 9 months via app or website. Understanding user engagement with these media is vital to developing successful interventions. Engagement is a complex, multifactorial concept that needs to move beyond simple metrics. Objective The aim of our study was to describe the development of an engagement index (EI) to monitor participant interaction with the Growing healthy app. The index included a number of subindices and cut-points to categorize engagement. Methods The Growing program was a feasibility study in which 300 mother-infant dyads were provided with an app which included 3 push notifications that was sent each week. Growing healthy participants completed surveys at 3 time points: baseline (T1) (infant age ≤3 months), infant aged 6 months (T2), and infant aged 9 months (T3). In addition, app usage data were captured from the app. The EI was adapted from the Web Analytics Demystified visitor EI. Our EI included 5 subindices: (1) click depth, (2) loyalty, (3) interaction, (4) recency, and (5) feedback. The overall EI summarized the subindices from date of registration through to 39 weeks (9 months) from the infant’s date of birth. Basic descriptive data analysis was performed on the metrics and components of the EI as well as the final EI score. Group comparisons used t tests, analysis of variance (ANOVA), Mann-Whitney, Kruskal-Wallis, and Spearman correlation tests as appropriate. Consideration of independent variables associated with the EI score were modeled using linear regression models. Results The overall EI mean score was 30.0% (SD 11.5%) with a range of 1.8% - 57.6%. The cut-points used for high engagement were scores greater than 37.1% and for poor engagement were scores less than 21.1%. Significant explanatory variables of the EI score included: parity (P=.005), system type including “app only” users or “both” app and email users (P<.001), recruitment method (P=.02), and baby age at recruitment (P=.005). Conclusions The EI provided a comprehensive understanding of participant behavior with the app over the 9-month period of the Growing healthy program. The use of the EI in this study demonstrates that rich and useful data can be collected and used to inform assessments of the strengths and weaknesses of the app and in turn inform future interventions.