Roberto Armenise

Personalized suggestions by means of Collaborative Filtering: A comparison of two different model-based techniques

Recommendation systems are commonly used for suggesting products or services. Among different existing techniques, Model-Based Collaborative Filtering (MBCF) approaches have been proven to address scalability and cold-starting problems that often arise. In this paper we investigate two MBCF algorithms: Self-Organizing Maps (SOM) for Collaborative Filtering and Item-based Fuzzy Clustering Collaborative Filtering (IFCCF). These two techniques have been selected because preliminary results have proven that when applied to the clustering of users or items the quality of the recommendation system increases with respect to the k-means. Within recommendation systems, no comparison of these two techniques exists. Therefore, our experimentation is aimed at comparing these two techniques by means of MovieLens and Jester dataset in order to provide a guideline for their implementation in the e-Commerce domain.

Improving accuracy of recommendation system by means of Item-based Fuzzy Clustering Collaborative Filtering

Predicting user preferences is a challenging task. Different approaches for recommending products to the users are proposed in literature and collaborative filtering has been proved to be one of the most successful techniques. Some issues related to the quality of recommendation and to computational aspects still arise (e.g., scalability and cold-start recommendations). In this paper, we propose an Item-based Fuzzy Clustering Collaborative Filtering (IFCCF) in order to ensure the benefits of a model-based technique improving the quality of suggestions. Experimentation led by predicting ratings of MovieLens and Jester users makes this promising and worth to be further investigated in a cross-domain dataset.

A generative solution for ATM cash management

Optimizing cash in automatic teller machines (ATM) is challenging due unpredictability of withdrawals, but profitable because of the large number of tellers. Generally ATM cash management and optimization is performed manually, according to corporate policies and personnel experience. A non-optimal cash upload can lead to poor service when cash demand is underestimated and to unnecessary costs when demand is overestimated. Therefore, finding the best match between cash stock and demand becomes crucial to improve. Recently, some authors attempted to optimize the cash by modeling and forecasting the demand. However, the high variance and non-stationarity of the underlying stochastic process can affect reliability of such an approach. In this paper we suggest the application of genetic algorithms as means for searching and generating optimal upload strategies, able at the same time to minimize the daily amount of stocked money and to assure cash dispensing service. Experimentation led at Poste Italiane S.p.A. makes this promising and worth to be further investigated.

A generative approach to product bundling in the e-Commerce domain

The product bundling problem is addressed by a Genetic Algorithm. We propose a generative approach in order to find the bundle of products that best satisfies user preferences and requirements and, at the same time, to guarantee the satisfaction of merchant needs such as the minimization of the dead stocks. The proposed approach succeeds in finding the optimal trade-off between different and conflicting constraints. Experimentation investigates algorithm convergence under several conditions, such as a different number of products in the bundle, increasing number of constraints, and different user requirements.

Optimization of menu layouts by means of genetic algorithms

Menu systems are key components in modern graphical user interfaces (GUIs), either for traditional desktop applications, or for the latest web applications. The design of interface layout must consider different aspects resulting in a trade-off between often conflicting requirements. This trade-off is aimed at making effective use of interfaces in order to meet user preferences and to conform to standard guidelines at the same time. Assuming we are able to quantify such a trade-off, the problem of finding a menu system able to maximize it figures as a combinatorial optimization problem. In this paper we investigate the application of genetic algorithms as a viable approach to identifying solutions that can be used as a starting point for further fine-tuning.

Searching optimal menu layouts by linear genetic programming

Designing effective menu systems is a key ingredient to usable graphical user interfaces. This task generally relies only on human ability in building hierarchical structures. However, trading off different and partially opposite guidelines, standards and practices is time consuming and can exceed human skills in problem solving. Recent advances are showing that this task can be addressed by generative approaches which exploit evolutionary algorithms as means for evolving different and unexpected solutions. The search of optimal solutions is made not trivial due to different alternatives which lead to local optima and constraints which can invalidate large sectors of the search space and make valid solutions sparse. This problem can be addressed by choosing an appropriate algorithm. In this paper we face the problem of searching optimal solutions by Linear Genetic Programming in particular, and we compare the solution to more conventional approaches based on simple genetic algorithms and genetic programming. Experimental results are discussed and compared to human-made solutions.

A Preliminary Experience in Optimizing the Layout of Web Pages by Genetic Algorithms to Fit Mobile Devices

Getting access to web content by mobile devices is becoming widespread. This poses the need of adapting content that have been designed for desktop application to being delivered on smaller displays. In this paper we investigate the application of a genetic algorithm as means for an automatic adaptation of existent pages to mobile device requirements, reporting preliminary results and outlining problems to be faced and solved in order to make this approach robust.

Web form page in mobile devices: Optimization of layout with a simple genetic algorithm

Filling out a form on mobile devices is generally harder than on other terminals, due to the reduced keyboard and display size, entailing a higher fatigue and limiting the user experience. A solution to this problem can be based on reducing the input effort required to the user by auto-completion, and re-organizing the fields in order to provide first those with a higher prediction power. In this paper we assume to be able to predict the user input and we optimize the fields layout aiming at reducing on average the input actions.

Modeling and predicting the user next input by Bayesian reasoning

We focus on enhancing the user experience by predicting entries when a form is filled, according to past interactions. The purpose of having a predictive model of form filling is to reduce the amount of time required to fill a form, and thus to reduce the fatigue and repetitiveness associated to this common task. Generally predictive models ignore the values entered by users in the other fields in the form, and just focus on the value getting entered at the current field. This is a limit to the model capabilities. Instead, we are aimed at predicting the sequence of entries in a form, instead of the value of single fields in isolation. This is done by means of inference over a Bayesian network, able to compute the a posteriori probability that remaining fields will assume certain values, given the set of values entered so far. The model structure and parameters can be learned from a dataset of past entries. The paper investigates computational and convergence issues under both the closed world assumption and the open world assumptions. As case study, we considered forms used for online payment of money order used at Poste Italiane, and we exploited this approach to prototype two different solutions for desktop and mobile applications. Results of experimentation with a user test group prove the proposed approach is able to provide an effective and appreciated support in filling a form.

A validation study regarding a generative approach in choosing appropriate colors for impaired users

In many circumstances, concepts, ideas and emotions are mainly conveyed by colors. Color vision disorders can heavily limit the user experience in accessing Information Society. Therefore, color vision impairments should be taken into account in order to make information and services accessible to a broader audience. The task is not easy for designers that generally are not affected by any color vision disorder. In any case, the design of accessible user interfaces should not lead to to boring color schemes. The selection of appealing and harmonic color combinations should be preserved. In past research we investigated a generative approach led by evolutionary computing in supporting interface designers to make colors accessible to impaired users. This approach has also been followed by other authors. The contribution of this paper is to provide an experimental validation to the claim that this approach is actually beneficial to designers and users.