Exploiting Latent Information in Recommender Systems

<p>This thesis exploits latent information in personalised recommendation, and investigates how this information can be used to improve recommender systems. The investigations span three directions: scalar rating-based collaborative filtering, distributional rating-based collaborative filtering, and distributional ratingbased hybrid filtering. In the first investigation, the thesis discovers through data analysis three problems in nearest neighbour collaborative filtering — item irrelevance, preference imbalance, and biased average — and identifies a solution: incorporating “target awareness” in the computation of user similarity and rating deviation. Two new algorithms are subsequently proposed. Quantitative experiments show that the new algorithms, especially the first one, are able to significantly improve the performance under normal situations. They do not however excel in cold-start situations due to greater demand of data. The second investigation builds upon the experimental analysis of the first investigation, and examines the use of discrete probabilistic distributional modelling throughout the recommendation process. It encompasses four ideas: 1) distributional input rating, which enables the explicit representation of noise patterns in user inputs; 2) distributional voting profile, which enables the preservation of not only shift but also spread and peaks in user’s rating habits; 3) distributional similarity, which enables the untangled and separated similarity computation of the likes and the dislikes; and 4) distributional prediction, which enables the communication of the uncertainty, granularity, and ambivalence in the recommendation results. Quantitative experiments show that this model is able to improve the effectiveness of recommendation compared to the scalar model and other published discrete probabilistic models, especially in terms of binary and list recommendation accuracy. The third investigation is based on an analysis regarding the relationship between rating, item content, item quality, and “intangibles”, and is enabled by the discrete probabilistic model proposed in the second investigation. Based on the analysis, a fundamentally different hybrid filtering structure is proposed, where the hybridisation strategy is neither linear nor sequential, but of a divide-and-conquer shape backed by probabilistic derivation. Experimental results show that it is able to outperform the standard linear and sequential hybridisation structures.</p>

Exploiting Latent Information in Recommender Systems

<p>This thesis exploits latent information in personalised recommendation, and investigates how this information can be used to improve recommender systems. The investigations span three directions: scalar rating-based collaborative filtering, distributional rating-based collaborative filtering, and distributional ratingbased hybrid filtering. In the first investigation, the thesis discovers through data analysis three problems in nearest neighbour collaborative filtering — item irrelevance, preference imbalance, and biased average — and identifies a solution: incorporating “target awareness” in the computation of user similarity and rating deviation. Two new algorithms are subsequently proposed. Quantitative experiments show that the new algorithms, especially the first one, are able to significantly improve the performance under normal situations. They do not however excel in cold-start situations due to greater demand of data. The second investigation builds upon the experimental analysis of the first investigation, and examines the use of discrete probabilistic distributional modelling throughout the recommendation process. It encompasses four ideas: 1) distributional input rating, which enables the explicit representation of noise patterns in user inputs; 2) distributional voting profile, which enables the preservation of not only shift but also spread and peaks in user’s rating habits; 3) distributional similarity, which enables the untangled and separated similarity computation of the likes and the dislikes; and 4) distributional prediction, which enables the communication of the uncertainty, granularity, and ambivalence in the recommendation results. Quantitative experiments show that this model is able to improve the effectiveness of recommendation compared to the scalar model and other published discrete probabilistic models, especially in terms of binary and list recommendation accuracy. The third investigation is based on an analysis regarding the relationship between rating, item content, item quality, and “intangibles”, and is enabled by the discrete probabilistic model proposed in the second investigation. Based on the analysis, a fundamentally different hybrid filtering structure is proposed, where the hybridisation strategy is neither linear nor sequential, but of a divide-and-conquer shape backed by probabilistic derivation. Experimental results show that it is able to outperform the standard linear and sequential hybridisation structures.</p>

An investigation into reading strategies used by Vietnamese non-English major students at Kien Giang University

This paper aimed at investigating the frequencies of reading strategies employed by Vietnamese non-English major students while taking a reading comprehension test at Kien Giang University. Data was collected through a questionnaire delivered to 117 sophomores who majored in Economics, Accounting and Construction at Kien Giang University. The results from the descriptive statistics showed that Vietnamese non-English majors were medium strategy users. Of the three types of reading strategies, cognitive strategies were the most frequently used, followed by metacognitive and support reading strategies. Out of 27 reading strategies, students reportedly used item number 14 “I read the text again for better understanding.” at the highest frequency while rating item number 2 “I determined what the type of the text is.” the least frequency.

Pengembangan Sistem Rekomendasi Tempat Pembuatan Kerajinan Tradisional Bali

Bali adalah salah satu destinasi tempat tujuan para wisatawan yang sangat berkembang saat ini. Banyak wisatawan datang ke Bali untuk mencari kerajinan tradisional Bali. Dari hasil data kuisioner, 56.7 % dari responden sangat tertarik dan 40.3 % dari responden tersebut tertarik ingin tahu dimana pembuatan kerajinan tradisional Bali. Wisatawan hanya mengetahui pusat pembelian kerajinan tanpa tahu dimana tempat pembuatannya. Sistem Rekomendasi dapat mengenalkan tempat pembuatan kerajinan tradisional Bali dengan membangun aplikasi. Sistem rekomendasi adalah sistem yang memberikan rekomendasi kepada pengguna dalam menemukan tempat pembuatan kerajinan tradisional bali berdasarkan pengguna sebelumnya. Sistem rekomendasi ini dibangun dengan menggunakan metode ICHM (item-basedclusteringhybridmethod) dan algoritmaslopeone. Dimana sistem ini akan memberikan suatu rekomendasi tempat dan kerajinan berdasarkan rating item dan konten item. Pengujian menggunakan MAE (MeanAverange Eror) pada sistem mendapatkan nilai kurang dari 1,000. Semakin rendah nilai MAE maka nilai rekomendasi semakin akurat.

A Compromise between Self-Enhancement and Honesty: Chinese Self-Evaluations on Social Desirability Scales

In this investigation, 202 Chinese college students were asked to complete the 48-item revised Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960) which contains 32 items from the original version for the purpose of rating item desirability and estimating the percentage of others in general who would behave in the manner described by these items. Analysis indicated (a) nearly all original items keyed in the original direction, which suggests similar fundamental values are prevalent among American and Chinese college students; (b) the distribution of Chinese scores on the 32 items was somewhat positively skewed rather than negatively skewed as in some Western studies of American and Canadian college students. Also, (c) Chinese subjects perceived that they did significantly more desirable and slightly fewer undesirable things than others from which one may infer that Chinese college students tended to give both self-enhancing and honest responses to present good images; however, their need for self-enhancement may take precedence over the need to be honest. (d) Subjects chose to give more honest responses to undesirable items than to desirable ones because the more undesirable items were rated as approximately more neutral than were more desirable ones. Hence, endorsing such undesirable items would not threaten their self-esteem or face. It can be seen that Chinese subjects made an intelligent compromise between self-enhancement and honesty.

Co-Operative Group versus Individual Evaluations of Instruction in Professional In-Service Training

School principals and senior school administrators evaluated an in-service training course in which they had participated, concerned with the use of co-operative learning programs in schools. Approximately half of the participants made their evaluations as individuals, whereas the remaining participants made their evaluations as members of three-person groups which were asked to reach consensus on each rating item. Course evaluations were more positive in the group-evaluation condition. The results are discussed in terms of group polarisation theory and validity issues concerning the uses of such evaluations in higher education settings.