Chapter 5 - Continuous Model Selection for Large-Scale Recommender Systems

While the prediction accuracy of a large-scale recommender system can generally be improved by learning from more and more training data over time, it is unclear how well a fixed predictive model can handle the changing business dynamics in a real-world scenario. The adjustment of a predictive model is controlled by the hyperparameter settings of a selected algorithm. Although the problem of hyperparameter optimization has been studied for decades in various disciplines, the adaptiveness of the initially selected model is not as well understood. This paper presents an approach to continuously re-select hyperparameter settings of the algorithm in a large-scale retail recommender system. In particular, an automatic hyperparameter optimization technique is applied on collaborative filtering algorithms in order to improve prediction accuracy. Experiments have been conducted on a large-scale real retail dataset to challenge traditional assumption that a one-off initial hyperparameter optimization is sufficient. The proposed approach has been compared with a baseline approach and a widely used approach with two scalable collaborative filtering algorithms. The evaluations of our experiments are based on a 2-year real purchase transaction dataset of a large retail chain business, both its online e-commerce site and its offline retail stores. It is demonstrated that continuous hyperparameter optimization can effectively improve the prediction accuracy of a recommender system. This paper presents a new direction in improving the prediction performance of a large-scale recommender system.

Recent years saw an explosive growth in the amount of digital information and the number of users who interact with this information through various platforms, ranging from web services to mobile applications and smart devices. This increase in information and users has naturally led to information overload which inherently limits the capacity of users to discover and find their needs among the staggering array of options available at any given time, the majority of which they may never become aware of. Online services have handled this information overload by using algorithmic filtering tools that can suggest relevant and personalized information to users. These filtering methods, known as Recommender Systems (RS), have become essential to recommend a range of relevant options in diverse domains ranging from friends, courses, music, and restaurants, to movies, books, and travel recommendations. Most research on recommender systems has focused on developing and evaluating models that can make predictions efficiently and accurately, without taking into account other desiderata such as fairness and transparency which are becoming increasingly important to establish trust with human users. For this reason, researchers have been recently pressed to develop recommendation systems that are endowed with the increased ability to explain why a recommendation is given, and hence help users make more informed decisions. Nowadays, state of the art Machine Learning (ML) techniques are being used to achieve unprecedented levels of accuracy in recommender systems. Unfortunately, most models are notorious for being black box models that cannot explain their output predictions. One such example is Matrix Factorization, a technique that is widely used in Collaborative Filtering algorithms. Unfortunately, like all black box machine learning models, MF is unable to explain its outputs. This dissertation proposes a new Cosine-based explainable Matrix Factorization model (CEMF) that incorporates a user-neighborhood explanation matrix (NSE) and incorporates a cosine based penalty in the objective function to encourage predictions that are explainable. Our evaluation experiments demonstrate that CEMF can recommend items that are more explainable and diverse compared to its competitive baselines, and that it further achieves this superior performance without sacrificing the accuracy of its predictions.

Recommender Systems (RS) aim at predicting items or ratings of items that the user are interested in. Collaborative Filtering (CF) algorithms such as user- and item-based methods are the dominant techniques applied in RS algorithms. To improve recommendation quality, metadata such as content information of items has typically been used as additional knowledge. With the increasing popularity of the collaborative tagging systems, tags could be interesting and useful information to enhance RS algorithms. Unlike attributes which are "global" descriptions of items, tags are "local" descriptions of items given by the users. To the best of our knowledge, there hasn't been any prior study on tag-aware RS. In this paper, we propose a generic method that allows tags to be incorporated to standard CF algorithms, by reducing the three-dimensional correlations to three two-dimensional correlations and then applying a fusion method to re-associate these correlations. Additionally, we investigate the effect of incorporating tags information to different CF algorithms. Empirical evaluations on three CF algorithms with real-life data set demonstrate that incorporating tags to our proposed approach provides promising and significant results.

The amount of movie has increased to become more congested; therefore, to find a movie what users are looking for through the existing technologies are very hard. For this reason, the users want a system that can suggest the movie requirement to them and the best technology about these is the recommendation system. However, the most recommendation system is using collaborative filtering methods to predict the needs of the user due to this method gives the most accurate prediction. Today, many researchers are paid attention to develop several methods to improve accuracy rather than using collaborative filtering methods. Hence, to further improve accuracy in the recommendation system, we present the k-clique methodology used to analyze social networks to be the guidance of this system. In this paper, we propose an efficient movie recommendation algorithm based on improved k-clique methods which are the best accuracy of the recommendation system. However, to evaluate the performance; collaborative filtering methods are monitored using the k nearest neighbors, the maximal clique methods, the k-clique methods, and the proposed methods are used to evaluate the MovieLens data. The performance results show that the proposed methods improve more accuracy of the movie recommendation system than any other methods used in this experiment.

<abstract><p>Collaborative filtering (CF) algorithm is one of the most widely used recommendation algorithms in recommender systems. However, there is a data sparsity problem in the traditional CF algorithm, which may reduce the recommended efficiency of recommender systems. This paper proposes an improved collaborative filtering personalized recommendation (ICF) algorithm, which can effectively improve the data sparsity problem by reducing item space. By using the k-means clustering method to secondarily extract the similarity information, ICF algorithm can obtain the similarity information of users more accurately, thus improving the accuracy of recommender systems. The experiments using MovieLens and Netflix data set show that the ICF algorithm has a significant improvement in the accuracy and quality of recommendation.</p></abstract>

This paper provides a detailed description of the recommendation system and collaborative filtering algorithm to optimize the English learning platform through the collaborative filtering algorithm and analyses the algorithmic principles and specific techniques of collaborative filtering. After introducing the recommendation system and collaborative filtering algorithm, this paper elaborates on the theoretical basis and technical principles of the recommendation algorithm based on cognitive ability and difficulty and provides an in‐depth analysis of the design and implementation of the recommendation algorithm by combining cognitive diagnosis theory, readability formula, and English knowledge map, which provides a comprehensive and solid theoretical guidance and support for the application development of the online English learning platform. The system is tested by building a Spring Cloud platform, importing actual business data, focusing on the validation of the recommendation model, and connecting the recommendation system to the formal production system to analyse the recommendation effect. Compared with the original recommendation method, the online English learning platform designed and implemented in this paper based on the cognitive ability and difficulty collaborative filtering recommendation algorithm has a better recommendation effect. The system is proved to be well designed and has certain reference and guiding value for the whole web‐based online learning platform and has a broader application prospect nowadays and in the future.

With the spread of COVID-19, the “untact” culture in South Korea is expanding and customers are increasingly seeking for online services. A recommendation system serves as a decision-making indicator that helps users by suggesting items to be purchased in the future by exploring the symmetry between multiple user activity characteristics. A plethora of approaches are employed by the scientific community to design recommendation systems, including collaborative filtering, stereotyping, and content-based filtering, etc. The current paradigm of recommendation systems favors collaborative filtering due to its significant potential to closely capture the interest of a user as compared to other approaches. The collaborative filtering harnesses features like user-profile details, visited pages, and click information to determine the interest of a user, thereby recommending the items that are related to the user’s interest. The existing collaborative filtering approaches exploit implicit and explicit features and report either good classification or prediction outcome. These systems fail to exhibit good results for both measures at the same time. We believe that avoiding the recommendation of those items that have already been purchased could contribute to overcoming the said issue. In this study, we present a collaborative filtering-based algorithm to tackle big data of user with symmetric purchasing order and repetitive purchased products. The proposed algorithm relies on combining extreme gradient boosting machine learning architecture with word2vec mechanism to explore the purchased products based on the click patterns of users. Our algorithm improves the accuracy of predicting the relevant products to be recommended to the customers that are likely to be bought. The results are evaluated on the dataset that contains click-based features of users from an online shopping mall in Jeju Island, South Korea. We have evaluated Mean Absolute Error, Mean Square Error, and Root Mean Square Error for our proposed methodology and also other machine learning algorithms. Our proposed model generated the least error rate and enhanced the prediction accuracy of the recommendation system compared to other traditional approaches.

Recommender systems are admittedly the widely used applications over the internet E commerce sites and thus the success of a recommender system depends on the time and accuracy of the results returned in response to information supplied by the users.Now-a-days, many big E commerce systems and even social networking portals provide the facility of recommendation on their sites, thus underscoring the demand for effective and accurate recommendation system. But still most of the recommender systems suffer from the problem of cold start, sparsity and popularity bias of the provided recommendation. Also, these recommender systems are unable to provide the recommendation to someone with unique taste. This paper describes a preference based recommender system and collaborative filtering approach which is used to solve prediction and recommendation problem. The collaborative filtering aims at learning predictive model of user interests, behavior from community data or user preferences. It describes a family of model-based approaches designed for this task. Probabilistic latent semantic analysis (PLSA) was presented in the context of data retrieval or text analysis area.This work can be used to predict user ratings in the recommendation system context. It is based on a statistical modeling technique which introduces a latent class of variables in a mixture model setting to discover prototypical interest profiles and user communities. The main advantages of PLSA over memory-based methods are an explicit, constant time prediction and more accurate compact model. It can be used to mine for the user community. The experimental results show substantial improvement in prediction accuracy over existing methods.

The inherent weakness of the data on user ratings collected from the web, such as sparsity and cold-start, has limited the data analysis capability and prediction accuracy in recommender systems. To alleviate this problem, trust has been incorporated in collaborative filtering approaches with encouraging experimental results. In this paper, we propose a computational model for trust-based CF with three different methods to infer trust in a social network, based on a detailed data analysis of hotel dataset. We apply these methods on users ratings of hotels and show its feasibility by comparing the testing results with conventional CF algorithm using evaluation metrics Mean absolute error (MAE) and prediction coverage. Our experimental results indicate that the use of trust can improve prediction accuracy if the definition of trust is reasonable enough.

Personalized recommendation systems can help people to find interesting things and they are widely used with the development of electronic commerce. Many recommendation systems employ the collaborative filtering technology, which has been proved to be one of the most successful techniques in recommender systems in recent years. With the gradual increase of users and items in electronic commerce systems, the time consuming nearest neighbor collaborative filtering search of the target customer in the total user space resulted in the failure of ensuring the real time requirement of recommender system. The paper proposed a personalized recommendation algorithm using hybrid information. User and item information is used for collaborative filtering to produce the recommendations. The recommendation joining user information and item information collaborative filtering is more scalable than the traditional one. Introduction The recommendation system in the electronic commerce website obtains the information from the customer, obtains the customer's interest, the hobby and the recommendation related commodity to the customer. Because the recommended system has great potential, so the research on the recommendation system is more and more. Collaborative filtering algorithm is one of the earliest and most successful recommendation techniques. At present, the algorithm is successful in both practice and research. However, with the deepening of the application and the wide range of applications, it also exposes some points, the most important is the following two points. The algorithm can be used in real time. Search for thousands of and target with similar interests, backgrounds of customers but the huge online customer groups, to find the speed of the neighbors to be improved, and thus improve the scalability of the algorithm. The recommended quality of the recommendation system is as easy to make two kinds of errors as other query system. However, the recommended system is not recommended for the mistake, that is, the customer does not like the item recommendation system is recommended. In electronic commerce, we should try to avoid this kind of mistake. Because, the error will cause the customer's dissatisfaction, thus destroys the recommendation system's authority. With the continuous expansion of the scale of electronic commerce system, the number of and the number of goods has increased dramatically, resulting in the calculation of the target user's nearest neighbor is not accurate, thereby reducing the system's recommendation quality. The paper proposed a personalized recommendation algorithm using hybrid information. User and item information is used for collaborative filtering to produce the recommendations. The recommendation joining user information and item information collaborative filtering is more scalable than the traditional one. 4th International Conference on Mechatronics, Materials, Chemistry and Computer Engineering (ICMMCCE 2015) © 2015. The authors Published by Atlantis Press 1434 Problem analysis Collaborative filtering is the key problem to find the nearest neighbor, this step is the key to the whole process. And to find the best neighbor, we have to effectively build a good user project score matrix. On the other hand, as customers, the continuous growth of goods, resulting in the user rating data is extremely sparse. The results are not recommended for certain and thus affect the accuracy of the recommendation. Based on the product features, the collaborative filtering recommendation is proposed to construct the product file, and the potential of the user to a certain kind of product is based on the user's purchase quantity and the average purchase amount. Because of the construction of collaborative filtering user project matrix. It is generally based on the user's score for all projects, as shown in figure 1. Figure 1 An example For example, you can consider in this regard, the project will be divided into different categories. Users in the same category of projects are more similar than in different types of projects more accurate. And it can also reduce the data's high dimension. Finally, the user's interest in each part of the project is calculated with a certain heuristic rule to merge each class of recommendation results as the user's final recommendation. Personalized collaborative recommendation algorithm based on hybrid information 1. The model of the method Collaborative filtering is one of the methods used in e-commerce recommendation system. Its core idea is that users tend to buy the goods which have similar interests. The basic starting point is that the customer is able to classify the customers. These three points constitute the basis of collaborative filtering, the more the user interest in the product, the higher the evaluation, as shown in figure 2.

Personal Learning Environment (PLE) is an e-learning concept that allows users to manage their learning environment both in terms of content and process. However, significant problems with PLE implementation in distance learning are excessive information and difficulties in finding the suitable learning content for learners. To overcome these problems, an experimental study was conducted to explore a learning content recommendation system for learners. The learning content recommendation system uses the Collaborative Filtering (CF) algorithm for the basis. CF is a method for filtering information by collecting ratings and combining it with similar information needs or interests of other users. This study intends to build the concept of PLE distance learning by applying the CF recommendation system to find learning content that is appropriate to the needs of learners. The test results reveal that the proposed PLE application is compliant with the PLE attributes. This study has also succeeded in applying a recommendation system using the CF algorithm with the concept of PLE in distance learning. Moreover, the Mean Absolute Error (MAE) calculation reveals that the best-obtained recommendation results reached by K=10. Based on the experimental data obtained, the greater the value of K used in the CF algorithm, the greater the average error.

As one of the most widely used algorithms in recommender systems, collaborative filtering algorithm faces two serious problems, which are the sparsity of data and poor recommendation effect in big data environment. In traditional clustering analysis, the object is strictly divided into several classes and the boundary of this division is very clear. However, for most objects in real life, there is no strict definition of their forms and attributes of their class. Concerning the problems above, this paper proposes to improve the traditional collaborative filtering model through the hybrid optimization of implicit semantic algorithm and fuzzy clustering algorithm, meanwhile, cooperating with collaborative filtering algorithm. In this paper, the fuzzy clustering algorithm is introduced to fuzzy clustering the information of project attribute, which makes the project belong to different project categories with different membership degrees, and increases the density of data, effectively reduces the sparsity of data, and solves the problem of low accuracy which is resulted from the inaccuracy of similarity calculation. Finally, this paper carries out empirical analysis on the MovieLens dataset, and compares it with the traditional user-based collaborative filtering algorithm. The proposed algorithm has greatly improved the recommendation accuracy.

Recommender system is the system which, by using automated information filtering technology, recommends products or services to the customers who are likely to be interested in. Those systems are widely used in many different Web retailers such as Amazon.com, Netfix.com, and CDNow.com. Various recommender systems have been developed. Among them, Collaborative Filtering (CF) has been known as the most successful and commonly used approach. CF identifies customers whose tastes are similar to those of a given customer, and recommends items those customers have liked in the past. Numerous CF algorithms have been developed to increase the performance of recommender systems. However, the relative performances of CF algorithms are known to be domain and data dependent. It is very time-consuming and expensive to implement and launce a CF recommender system, and also the system unsuited for the given domain provides customers with poor quality recommendations that make them easily annoyed. Therefore, predicting in advance whether the performance of CF recommender system is acceptable or not is practically important and needed. In this study, we propose a decision making guideline which helps decide whether CF is adoptable for a given application with certain transaction data characteristics. Several previous studies reported that sparsity, gray sheep, cold-start, coverage, and serendipity could affect the performance of CF, but the theoretical and empirical justification of such factors is lacking. Recently there are many studies paying attention to Social Network Analysis (SNA) as a method to analyze social relationships among people. SNA is a method to measure and visualize the linkage structure and status focusing on interaction among objects within communication group. CF analyzes the similarity among previous ratings or purchases of each customer, finds the relationships among the customers who have similarities, and then uses the relationships for recommendations. Thus CF can be modeled as a social network in which customers are nodes and purchase relationships between customers are links. Under the assumption that SNA could facilitate an exploration of the topological properties of the network structure that are implicit in transaction data for CF recommendations, we focus on density, clustering coefficient, and centralization which are ones of the most commonly used measures to capture topological properties of the social network structure. While network density, expressed as a proportion of the maximum possible number of links, captures the density of the whole network, the clustering coefficient captures the degree to which the overall network contains localized pockets of dense connectivity. Centralization reflects the extent to which connections are concentrated in a small number of nodes rather than distributed equally among all nodes. We explore how these SNA measures affect the performance of CF performance and how they interact to each other. Our experiments used sales transaction data from H department store, one of the well?known department stores in Korea. Total 396 data set were sampled to construct various types of social networks. The dependant variable measuring process consists of three steps; analysis of customer similarities, construction of a social network, and analysis of social network patterns. We used UCINET 6.0 for SNA. The experiments conducted the 3-way ANOVA which employs three SNA measures as dependant variables, and the recommendation accuracy measured by F1-measure as an independent variable. The experiments report that 1) each of three SNA measures affects the recommendation accuracy, 2) the density's effect to the performance overrides those of clustering coefficient and centralization (i.e., CF adoption is not a good decision if the density is low), and 3) however though the density is low, the performance of CF is comparatively good when the clustering coefficient is low. We expect that these experiment results help firms decide whether CF recommender system is adoptable for their business domain with certain transaction data characteristics.

The proliferation of powerful smart devices is revolutionizing mobile computing systems. A particular set of applications that is gaining wide interest is recommender systems. Recommender systems provide their users with recommendations on variety of personal and relevant items or activities. They can play a significant role in today's life whether in E-commerce or for daily decisions that we need to make. We introduce a hybrid approach for solving the problem of finding the ratings of unrated items in a user-item ranking matrix through a weighted combination of user-based and item-based collaborative filtering. The proposed technique provides improvements in addressing two major challenges of recommender systems: accuracy of recommender systems and sparsity of data by simultaneously incorporating users' correlations and items ones. The evaluation of the system shows superiority of the solution compared to stand-alone user-based collaborative filtering or item-based collaborative filtering.

PurposeImproving the performance of recommender systems (RSs) has always been a major challenge in the area of e-commerce because the systems face issues such as cold start, sparsity, scalability and interest drift that affect their performance. Despite the efforts made to solve these problems, there is still no RS that can solve or reduce all the problems simultaneously. Therefore, the purpose of this study is to provide an effective and comprehensive RS to solve or reduce all of the above issues, which uses a combination of basic customer information as well as big data techniques.Design/methodology/approachThe most important steps in the proposed RS are: (1) collecting demographic and behavioral data of customers from an e-clothing store; (2) assessing customer personality traits; (3) creating a new user-item matrix based on customer/user interest; (4) calculating the similarity between customers with efficient k-nearest neighbor (EKNN) algorithm based on locality-sensitive hashing (LSH) approach and (5) defining a new similarity function based on a combination of personality traits, demographic characteristics and time-based purchasing behavior that are the key incentives for customers' purchases.FindingsThe proposed method was compared with different baselines (matrix factorization and ensemble). The results showed that the proposed method in terms of all evaluation measures led to a significant improvement in traditional collaborative filtering (CF) performance, and with a significant difference (more than 40%), performed better than all baselines. According to the results, we find that our proposed method, which uses a combination of personality information and demographics, as well as tracking the recent interests and needs of the customer with the LSH approach, helps to improve the effectiveness of the recommendations more than the baselines. This is due to the fact that this method, which uses the above information in conjunction with the LSH technique, is more effective and more accurate in solving problems of cold start, scalability, sparsity and interest drift.Research limitations/implicationsThe research data were limited to only one e-clothing store.Practical implicationsIn order to achieve an accurate and real-time RS in e-commerce, it is essential to use a combination of customer information with efficient techniques. In this regard, according to the results of the research, the use of personality traits and demographic characteristics lead to a more accurate knowledge of customers' interests and thus better identification of similar customers. Therefore, this information should be considered as a solution to reduce the problems of cold start and sparsity. Also, a better judgment can be made about customers' interests by considering their recent purchases; therefore, in order to solve the problems of interest drifts, different weights should be assigned to purchases and launch time of products/items at different times (the more recent, the more weight). Finally, the LSH technique is used to increase the RS scalability in e-commerce. In total, a combination of personality traits, demographics and customer purchasing behavior over time with the LSH technique should be used to achieve an ideal RS. Using the RS proposed in this research, it is possible to create a comfortable and enjoyable shopping experience for customers by providing real-time recommendations that match customers' preferences and can result in an increase in the profitability of e-shops.Originality/valueIn this study, by considering a combination of personality traits, demographic characteristics and time-based purchasing behavior of customers along with the LSH technique, we were able for the first time to simultaneously solve the basic problems of CF, namely cold start, scalability, sparsity and interest drift, which led to a decrease in significant errors of recommendations and an increase in the accuracy of CF. The average errors of the recommendations provided to users based on the proposed model is only about 13%, and the accuracy and compliance of these recommendations with the interests of customers is about 92%. In addition, a 40% difference between the accuracy of the proposed method and the traditional CF method has been observed. This level of accuracy in RSs is very significant and special, which is certainly welcomed by e-business owners. This is also a new scientific finding that is very useful for programmers, users and researchers. In general, the main contributions of this research are: 1) proposing an accurate RS using personality traits, demographic characteristics and time-based purchasing behavior; 2) proposing an effective and comprehensive RS for a “clothing” online store; 3) improving the RS performance by solving the cold start issue using personality traits and demographic characteristics; 4) improving the scalability issue in RS through efficient k-nearest neighbors; 5) Mitigating the sparsity issue by using personality traits and demographic characteristics and also by densifying the user-item matrix and 6) improving the RS accuracy by solving the interest drift issue through developing a time-based user-item matrix.