Pairwise Ranking Research Articles

Optimal nonparametric estimator of the area under ROC curve based on clustered data

In diagnostic trials, clustered data are obtained when several subunits of the same patient are observed. Intracluster correlations need to be taken into account when analyzing such clustered data. A nonparametric method has been proposed by Obuchowski (1997) to estimate the Receiver Operating Characteristic curve area (AUC) for such clustered data. However, Obuchowski’s estimator is not efficient as it gives equal weight to all pairwise rankings within and between cluster. In this paper, we propose a more efficient nonparametric AUC estimator with two sets of optimal weights. Simulation results show that the loss of efficiency of Obuchowski’s estimator for a single AUC or the AUC difference can be substantial when there is a moderate intracluster test correlation and the cluster size is large. The efficiency gain of our weighted AUC estimator for a single AUC or the AUC difference is further illustrated using the data from a study of screening tests for neonatal hearing.

Joint embeddings with multimodal cues for video-text retrieval

For multimedia applications, constructing a joint representation that could carry information for multiple modalities could be very conducive for downstream use cases. In this paper, we study how to effectively utilize available multimodal cues from videos in learning joint representations for the cross-modal video-text retrieval task. Existing hand-labeled video-text datasets are often very limited by their size considering the enormous amount of diversity the visual world contains. This makes it extremely difficult to develop a robust video-text retrieval system based on deep neural network models. In this regard, we propose a framework that simultaneously utilizes multimodal visual cues by a “mixture of experts” approach for retrieval. We conduct extensive experiments to verify that our system is able to boost the performance of the retrieval task compared to the state of the art. In addition, we propose a modified pairwise ranking loss function in training the embedding and study the effect of various loss functions. Experiments on two benchmark datasets show that our approach yields significant gain compared to the state of the art.

USRRM: Pairwise Ranking and Scoring Images Using Its Aesthetic Quality

Image Aesthetics Analysis is a challenging research problem as aesthetics of an image is a subjective quality and it is quite difficult to formulate it into a mathematical or algorithmic problem. On the other hand, its applications are numerous, ranging from aesthetic based image retrieval to image editing. Earlier works on image aesthetics analysis relied upon handpicking the standard features from the image, based upon which, its aesthetics was quantified. This method was applied, in the belief that sufficient aesthetic features have been taken into consideration and no more features impact its aesthetic quality. This is not always the case, since, subjective quality as this, is defined individually and depends upon personal perspective. With the advent of deep learning, automatic feature learning became prevalent. The classification works on image aesthetics using deep learning have had some success. However, we are concerned with giving a tentative score to images and perform some sort of relative ranking among them. We formulate a unified loss objective accounting both of these factors and also devise a model named USRRM which learns from the global view and pixel-level finer details from an image. The external style and semantic information also aid in model learning. The scoring result of USRRM has the best correlation with the ground truth scores among the most similar previous models and comparable results with less similar models but evaluated with the same metrics. We quantify our results using different correlation coefficients and an accuracy metric.

Land Suitability Assessment for Potato Crop using Analytic Hierarchy Process Technique and Geographic Information System

Agricultural land-suitability analysis is a prerequisite to achieve optimum utilization of the available land resources for sustainable agricultural production. The Analytical Hierarchy Process (AHP) technique coupled with Geographic Information System (GIS) can be a unique tool for land-suitability studies. AHP-GIS technique based on soil nutrient criteria of soil texture, pH, organic carbon, electric conductivity, available nitrogen, phosphorous, potassium, and zinc was used for land-suitability assessment in Tarkeswar Block, Hooghly district, West Bengal for growing potato. The study area was classified into suitable land categories based on soil nutrient levels analysed from 50 randomly-selected plot-based soil samples. Pair-wise comparison matrix-based ranking was computed considering the importance of each criterion for potato crop in the area. Suitability maps were developed and analysed in ArcGIS software environment. The total area was classified into ‘highly suitable’ class occupying 61 ha (21%), ‘moderately suitable’ class in 195 ha (67%), ‘marginally suitable’ class in 32 ha (11%) and ‘not suitable’ class in 3 ha (1%) land areas. Yield distribution map of potato crop showed nearly 106 ha (36.5%) area producing higher tuber yield of 20 t.ha-1. The proposed suitability map was validated against crop yield map where nearly 253 ha (87%) area classified under ‘highly suitable’ to ‘moderately suitable’ classes was found to give better potato yield of 15 t.ha-1or more. The AHP-GIS technique can be used for crop rotation analysis for multiple seasons for better crop selection in long run.

Sampler Design for Bayesian Personalized Ranking by Leveraging View Data

Bayesian Personalized Ranking (BPR) is a representative pairwise learning method for optimizing recommendation models. It is widely known that the performance of BPR depends largely on the quality of negative sampler. In this paper, we make two contributions with respect to BPR. First, we find that sampling negative items from the whole space is unnecessary and may even degrade the performance. Second, focusing on the purchase feedback of E-commerce, we propose a negative sampler for BPR by leveraging the additional view data. In our proposed sampler, users' viewed interactions are considered as an intermediate feedback between the purchased and unobserved interactions. We jointly learn the pairwise rankings of user preference among these three types of interactions and design a user-oriented weighting strategy during learning process, which is more effective and flexible. Compared to the vanilla BPR that applies a uniform sampler on all candidates, our view-enhanced sampler enhances BPR with a relative improvement over 36.64 and 16.40 percent on Beibei and Tmall datasets, respectively. Empirical studies demonstrate the importance of considering users' additional feedback when modeling their preference on different items, which can effectively improve the quality of sampled negative items towards learning a better personalized ranking function. Our implementation is available at https://github.com/dingjingtao/NegativeSamplerBPR.

Floristic Inventory and Ethnobotany of Wild Edible Plants in Cebu Island, Philippines

The study discussed the  diversity, utilization, socio-economic value and threats associated to the wild edible plants (WEPs) in Cebu Island, Philippines. Semi-structured  interview, field observations and pairwise ranking were utilized to  generate responses from purposively-selected  informants living in  three mountain communities in Cebu: Mt. Kapayas, Catmon, Nug-as Forest Reserve, Alcoy and Cantabaco Forest, Toledo. The study recorded a total of 168 edible plant species belonging to 111 genera and 97 families. Majority of plant habits of the recorded plants were trees (42.94%), herbs (30%) and shrubs (14.71%). The recorded use of WEPs include food, medicine, construction/technology, beverage, forage, food coloring and condiments where 30% of the informants suggest the marketability of the WEPs and 14.7% of the WEPs were reported as alternative food source. Test for use diversity showed significant difference in the uses of WEPs in the three sites (X2 = 2.46, α = 0.05). Pairwise-ranking results indicated that selective cutting for construction and technology, natural disaster and agricultural expansion are key threats to WEPs in these areas. The study suggests 1) public awareness and community-based management at all levels to lessen threats, 2) further investigation on nutritive and socio-economic values;and 3) investigation of pharmacological profiles to evaluate medicinal benefits.  Keywords : wild edible plants, ethnobotany, alternative food source

Factors Influencing Soil and Water Conservation Adoption in Basheri, Gheshnashm and Shmangus Laelai, Eritrea

<p>In response to widespread land degradation remarkable Soil and Water Conservation (SWC) interventions were carried out in Eritrea. Nonetheless, such interventions were not adopted by farmers and the problem still persists, hence this research was carried out from June 2015 to June 2016 to identify the factors affecting the SWC practices in Eritrea. Seventy two farm household heads were selected through quota simple random sampling and data was collected through group discussions, pair-wise problem ranking, participatory transect walks and semi-structured household questionnaire surveys. Descriptive statistics and binary logistic regression model were used for data analysis. The results indicated that only 23.61% of the respondents practice proper SWC in their farms. Moreover, the logistic regression analysis revealed that off-farm activities, land tenure insecurity, educational level of household head and household head being female had significant negative effects on SWC efforts, whereas slope of the farm showed significant positive effect. Land tenure insecurity, weak extension service and low agricultural production were also ranked by farmers as the first three constraints in SWC practices. As the legacy of the past food for work and cash for work programs, SWC is considered as a paid-work by local communities. Hence, appropriate agricultural extension, land tenure security and introduction of agroforestry farming systems are recommended in the study area.</p>

English

Bread wheat is one of the most important cereal crops of the world and a staple food for about one third of the world’s population and is a major cereal crop in Ethiopia. One of the major challenges in improving food security is to develop varieties that are adapted to specific environment and farmers’ needs. Field trials were conducted at two locations, Hitosa (Sero-Anketo kebele) and Limu Bilbilo (Bekoji-Negesso kebele) districts, in Arsi zone of Oromiya regional state, Ethiopia in 2015. The objectives were to identify farmers’ and traders’ preferences and selection criteria and acceptable varieties among the tasted twenty-five bread wheat varieties through farmers’ participation. The experiment was laid out in lattice design with three replications in which farmers participated only in one of the replication for ranking. Farmers and traders identified top seven criteria that are the same  at both locations (that is, disease and insect resistance, grain yield, spike size, seed color, tillering capacity, market demand and seed size, except seed weight instead of seed size at Seru-Anketo) for rating of varieties from 1 to 5 scale (1=very poor and 5=excellent). Data analysis was done using SAS and Microsoft Excel. All varieties showed resistant type of infection for the three rusts (Stem, Yellow and Leaf) at Bekoji-Negesso: As all varieties scored <20 ACI. Similarly, at Sero-Anketo, Kakaba, Digelu and Jefferson ranged under MS to S whereas Gassay, Hiddasse, and Mekelle-02 ranged under MS and MR types of infection for SR, respectively. Grain protein was analyzed and Hoggana (14.27%) was found to be the highest. Based on measured trait (rusts resistance) and farmers’ and traders preferences ranking; Bika, Bulluk and TAY for Bekoji-Negeso and Mekelle-4, Ogolcho and TAY for Sero-Anketo were recommended with their full production packages. Therefore, participation of farmers in early breeding program could be one of the approaches as to identify the best variety for specific location. Key words: Direct matrix ranking, grain protein, pairwise ranking matrix, participatory varietal selection (PVS), rust. &nbsp

Preferences for physical activity: a conjoint analysis involving people with chronic knee pain

To investigate individual preferences for physical activity (PA) attributes in adults with chronic knee pain, to identify clusters of individuals with similar preferences, and to identify whether individuals in these clusters differ by their demographic and health characteristics. An adaptive conjoint analysis (ACA) was conducted using the Potentially All Pairwise RanKings of all possible Alternatives (PAPRIKA) method to determine preference weights representing the relative importance of six PA attributes. Cluster analysis was performed to identify clusters of participants with similar weights. Chi-square and ANOVA were used to assess differences in individual characteristics by cluster. Multinomial logistic regression was used to assess associations between individual characteristics and cluster assignment. The study sample included 146 participants; mean age 65, 72% female, 47% white, non-Hispanic. The six attributes (mean weights in parentheses) are: health benefit (0.26), enjoyment (0.24), convenience (0.16), financial cost (0.13), effort (0.11) and time cost (0.10). Three clusters were identified: Cluster 1 (n=33): for whom enjoyment (0.35) is twice as important as health benefit; Cluster 2 (n=63): for whom health benefit (0.38) is most important; and Cluster 3 (n=50): for whom cost (0.18), effort (0.18), health benefit (0.17) and enjoyment (0.18) are equally important. Cluster 1 was healthiest, Cluster 2 most self-efficacious, and Cluster 3 was in poorest health. Patients with chronic knee pain have preferences for PA that can be distinguished effectively using ACA methods. Adults with chronic knee pain, clustered by PA preferences, share distinguishing characteristics. Understanding preferences may help clinicians and researchers to better tailor PA interventions.

APL: Adversarial Pairwise Learning for Recommender Systems

The main objective of recommender systems is to help users select their desired items, where a major challenge is modeling users’ preferences based on their historical feedback (e.g., clicks, purchases or check-ins). Recently, several recommendation models have utilized the adversarial technique, which has been successfully used to capture real data distributions in various domains (e.g., computer vision). Nevertheless, the training process of the original adversarial technique is very slow and unstable in the domain of recommender systems. First, the sparsity of the implicit feedback dataset aggravates the inherently intractable adversarial training process. Second, since the original adversarial model is designed for differentiable values (e.g., images), the discrete items also increase the training difficulty. To cope with these issues, we propose a novel method named Adversarial Pairwise Learning (APL), which unifies generative and discriminative models via adversarial learning. Specifically, based on the weaker assumption that the user prefers observed items over generated items, APL exploits pairwise ranking to accelerate the convergence and enhance the stability of adversarial learning. Additionally, a differentiable procedure is adopted to replace the discrete item sampling to optimize APL via backpropagation and stabilize the training process. Extensive experiments under multiple recommendation scenarios demonstrate APL’s effectiveness, fast convergence and stability. Our implementation of APL is available at: https://github.com/ZhongchuanSun/APL.

Analysis of cutaneous Merkel cell carcinoma outcomes after different surgical interventions

Current guidelines recommend local excision margin (EM) with 1 to 2cm on primary Merkel cell carcinoma (MCC) sites. We compared survival outcomes of patients with MCC who were treated with different surgical interventions. A retrospective analysis of MCC cases in the Surveillance, Epidemiology, and End Results database was performed using the Kaplan-Meier, competing risk, and Cox proportional hazards regression model analyses. Influence of age, T stage, American Joint Committee on Cancer stage, adjuvant radiotherapy, and other subgroups were also analyzed by pairwise log rank test. Our results indicated a significant association between local destruction method and inferior survival, while an EM >2cm showed significantly higher overall survival. In addition, the competing risk analysis depicted a similar trend as the Kaplan-Meier analysis, and considerably reduced estimated cumulative incidence. Further subgroup pairwise analysis demonstrated that the EM >2cm method had better survival in patients who were <60years of age, having smaller tumor diameters (T1 and T2) or having undergone adjuvant radiotherapy (P<.05). In contrast, different EMs did not show any significant association with survival rate in patients ≥75years of age or stage III tumors. This study was not prospectively randomized without relapse data. It is challenging to make optimal EM recommendations, because surgical options may depend on individual case situations. Further prospective randomized studies are warranted.

Clinical Decision-Making for Thrombolysis of Acute Minor Stroke Using Adaptive Conjoint Analysis.

There is practice variability in the treatment of patients with minor ischemic stroke with thrombolysis. We sought to determine which clinical factors physicians prioritize in thrombolysis decision-making for minor stroke using adaptive conjoint analysis. We conducted our conjoint analysis using the Potentially All Pairwise RanKings of all possible Alternatives methodology via the 1000Minds platform to design an online preference survey and circulated it to US physicians involved in stroke care. We evaluated 6 clinical attributes: language/speech deficits, motor deficits, other neurological deficits, history suggestive of increased risk of complication from thrombolysis, age, and premorbid disability. Survey participants were asked to choose between pairs of treatment scenarios with various clinical attributes; scenarios automatically adapted based on participants' prior responses. Preference weights representing the relative importance of each attribute were compared using unadjusted paired t tests. Statistical significance was set at α = .05. Fifty-four participants completed the survey; 61% were vascular neurologists and 93% worked in academic centers. All neurological deficits were ranked higher than age, premorbid status, or potential contraindications to thrombolysis. Differences between each successive mean preference weight were significant: motor (31.7%, standard deviation [SD]: 9.5), language/speech (24.1%, SD: 9.6), other neurological deficits (16.6%, SD: 6.4), premorbid status (12.9%, SD: 6.6), age (10.1%, SD: 6.3), and potential thrombolysis contraindication (4.7%, SD: 4.4). In a conjoint analysis, surveyed US physicians in academic practice assigned greater weight to motor and speech/language deficits than other neurological deficits, patient age, relative contraindications to thrombolysis, and premorbid disability when deciding to thrombolyse patients with minor stroke.

Entity representation for pairwise collaborative ranking using restricted Boltzmann machine

Pairwise preference data is one of several kinds of feedback data that many modern intelligent systems gather and process in order to help people make better decisions. Recommender systems are of main categories of such intelligent systems. An emerging recommendation approach in this domain is Pairwise Collaborative Ranking (PCR) which seeks to learn pairwise preferences of users over items. Current PCR algorithms generally learn latent factors for users and items that are consistent to the ground truth pairwise preferences, instead of directly extracting latent factors for pairwise preferences. Although this is the typical approach for rating oriented recommendation, it does not seem to be the best option for pairwise collaborative ranking. In this paper RBMRank, a novel model-based PCR approach, is introduced, that is designed for recommendation based on pairwise preference data. RBMRank uses the power of the restricted Boltzmann machines to map the entities of the system into dense feature spaces. Its embedding system is able to generate useful representations even when dealing with sparse data sets, that is a very important issue in recommender systems domain. The other important feature of RBMRank is that it is able to extract latent factors for pairwise preferences, instead of items, that allows it to directly predict the unknown preferences of each user. Experimental results on MovieLens datasets show up to 5.9% improvement of accuracy in terms of NDCG@5 and NDCG@10 compared to other state-of-the-art algorithms in the field of collaborative ranking.

Factors Influencing Cane Use for the Management of Knee Osteoarthritis: A Cross-Sectional Survey.

To investigate demographic, symptom-related, and cognitive determinants of cane use for knee osteoarthritis (OA) and prioritize the factors that could facilitate cane use in people with no previous cane use. A survey of people ages ≥45 years with a clinical diagnosis of knee OA was conducted. The survey consisted of the following two sections: 1) demographic and cognitive determinants of cane use assessed via subscales of the Cane Cognitive Mediator Scale, and 2) 19 statements, underpinned by the Behaviour Change Wheel theoretical framework, relating to factors that could facilitate regular cane use. Logistic regression was used to examine determinants of cane use, while a priority pairwise ranking activity (1000minds software) determined the rank order of the 19 statements that could facilitate cane use. A total of 529 people completed Part 1 (80% females; 35% cane users) and 231 people completed Part 2. Age (odds ratio [OR] 1.06, 95% confidence interval [95% CI] 1.03- 1.09), body mass index (BMI) (OR 1.03, 95% CI 1.01-1.06), knee pain ≥3 years (OR 2.62, 95% CI 1.63-4.21) and numeric rating scale pain level while walking (OR 1.19, 95% CI 1.09-1.30) were significant independent determinants of cane use. In people who had never used a cane, statements relating to cane-use technique, fitting, knowledge of benefits, and motivation were ranked highest overall. Independent determinants of cane use include older age, higher BMI, greater pain duration, and greater severity of knee pain. Strategies targeting an individual's capability and motivation to use a cane may increase cane use among people with knee OA.

Exploiting Web Images for Video Highlight Detection With Triplet Deep Ranking

Highlight detection from videos has been widely studied due to the fast growth of video contents. However, most existing approaches to highlight detection, either handcraft feature based or deep learning based, heavily rely on human-curated training data, which is very expensive to obtain and, thus, hinders the scalability to large datasets and unlabeled video categories. We observe that the largely available Web images can be applied as a weak supervision for highlight detection. For example, the top-ranked images in reference to the query “skiing” returned by a search engine may contain considerable positive samples of “skiing” highlights. Motivated by this observation, we propose a novel triplet deep ranking approach to video highlight detection using Web images as a weak supervision. The approach handles the relative preference of highlight scores between highlighting frames, nonhighlighting frames, and Web images by the triplet ranking constraints. Our approach can iteratively train two interdependent deep models (i.e., a triplet highlight model and a pairwise noise model) to deal with the noisy Web images in a single framework. We train the two models with relative preferences to generalize the capability regardless of the categories of training data. Therefore, our approach is fully category independent and exploits weakly supervised Web images. We evaluate our approach on two challenging datasets and achieve impressive results compared with the state-of-the-art pairwise ranking support vector machines, a robust recurrent autoencoder, and spatial deep convolution neural networks. We also empirically verify through cross-dataset evaluation that our category-independent model is fairly generalizable even if two different datasets do not share exactly the same categories.

Characterization of the Row Geometric Mean Ranking with a Group Consensus Axiom

An axiomatic approach is applied to the problem of extracting a ranking of the alternatives from a pairwise comparison ratio matrix. The ordering induced by row geometric mean method is proved to be uniquely determined by three independent axioms, anonymity (independence of the labelling of alternatives), responsiveness (a kind of monotonicity property) and aggregation invariance, which requires the preservation of group consensus, that is, the pairwise ranking between two alternatives should remain unchanged if unanimous individual preferences are combined by geometric mean.

On linear wine score aggregators and the preservation of group preferences

PurposeThis paper aims to identify the potential conflicts that arise between the actual and the revealed preference of a panel of wine judges when the panel’s evaluation is derived by a linear aggregation of individual scores.Design/methodology/approachA standard axiomatic social choice theoretical model is used to derive and examine the findings.FindingsThe findings show that even with the application of a simple majority rule over the pairwise ranking of wines, preferences may be misrepresented by the ordinal ranking of the wine score aggregation.Originality/valueA number of wine competitions and reviews, to date, use some form of linear aggregation to represent group preferences. Furthermore, tests surrounding wine judge performance are largely dependent on some underlying true measures usually derived from a linear aggregation. The results imply that care should be taken in these regards.

Community Perception to Reduce Safe Drinking Water Scarcity A Case Study at Debishing Para In Rajshahi City Corporation Area, Bangladesh

The safe drinking water crisis is recognized as one of the major problems and the significant barrier to improve health and wellbeing in Rajshahi city. Basically, RWASA (Rajshahi Water Supply and Sewerage Authority) supplies water at Debishingpara in Rajshahi City Corporation (RCC). Safe drinking water crisis is becoming intensified day by day in Rajshahi City. The objective of the study is to identify local problem regarding safe drinking water scarcity in the community and to propose a communitybased Strategies for reducing safe drinking water crisis. Participatory Rural Appraisal (PRA) approach was applied to identify the problems and suggest the community- based solution by involving community people. The study has been conducted by collecting data from a focus group which consists of 8 members (4 males and 4 females) on supplied water and water collection system. Causeeffect diagram and Pair-wise Ranking Matrix was used as PRA tools to understand the causes and impact of drinking water scarcity and identify the main difficulties to access safe drinking water. According to the focus group discussion, the community people suffered mostly due to lack of water purification system, less number of tube well and existence of iron in the water. The study includes a dream map which obtains suggestions from the participants of the focus group. This study emphasizes on the community participation in the participatory forum to discuss appropriate solutions relating to the supplied water in the community. This study also tries to incorporate relevant policy and strategy to improve the condition.

Discovering a taste for the unusual: exceptional models for preference mining

Exceptional preferences mining (EPM) is a crossover between two subfields of data mining: local pattern mining and preference learning. EPM can be seen as a local pattern mining task that finds subsets of observations where some preference relations between labels significantly deviate from the norm. It is a variant of subgroup discovery, with rankings of labels as the target concept. We employ several quality measures that highlight subgroups featuring exceptional preferences, where the focus of what constitutes ‘exceptional’ varies with the quality measure: two measures look for exceptional overall ranking behavior, one measure indicates whether a particular label stands out from the rest, and a fourth measure highlights subgroups with unusual pairwise label ranking behavior. We explore a few datasets and compare with existing techniques. The results confirm that the new task EPM can deliver interesting knowledge.

Exploiting geo-social correlations to improve pairwise ranking for point-of-interest recommendation

Recently, as location-based social network (LBSN) rapidly grow, point-of-in-terest (POI) recommendation has become an important way to help people locate interesting places. Nowadays, there have been deep studies conducted on the geographical and social influence in the point-of-interest recommendation model based on the rating prediction. The fact is, however, relying solely on the rating fails to reflect the user's preferences very accurately, because the users are most concerned with the list of ranked point-of-in-terests (POIs) on the actual output of recom-mender systems. In this paper, we propose a co-pairwise ranking model called Geo-Social Bayesian Personalized Ranking model (GSB-PR), which is based on the pairwise ranking with the exploiting geo-social correlations by incorporating the method of ranking learning into the process of POI recommendation. In this model, we develop a novel BPR pairwise ranking assumption by injecting users' geo-social preference. Based on this assumption, the POI recommendation model is reformulated by a three-level joint pairwise ranking model. And the experimental results based on real datasets show that the proposed method in this paper enjoys better recommendation performance compared to other state-of-the-art POI recommendation models.