Feature selection is an important research area that seeks to eliminate unwanted features from da... more Feature selection is an important research area that seeks to eliminate unwanted features from datasets. Many feature selection methods are suggested in the literature, but the evaluation of the best set of features is usually performed using supervised metrics, where labels are required. In this work we propose a methodology that tries to aid data specialists to answer simple but important questions, such as: (1) do current feature selection methods give similar results? (2) is there is a consistently better method ? (3) how to select the m-best features? (4) as the methods are not parameter-free, how to choose the best parameters in the unsupervised scenario? and (5) given different options of selection, could we get better results if we fusion the results of the methods? If yes, how can we combine the results? We analyze these issues and propose a methodology that, based on some unsupervised methods, will make feature selection using strategies that turn the execution of the proc...
ABSTRACT Generalized additive independent (GAI) preference models are, in many situations, more i... more ABSTRACT Generalized additive independent (GAI) preference models are, in many situations, more interesting than additive models, as GAI models allow interdepen-dencies between attributes. However, they are more difficult to construct (elicit), not only because the number of questions needed to be asked increases, but also because we must know what the interdependencies are, i.e. the structure of the model. We introduce the use of association rules to select attributes and detect simple interactions between them during an interactive preference elicitation pro-cess that intends to build a GAI utility function for the preferences of the user of a recommender system. Using this strategy, we have built a recommender system pro-totype that suggests touristic sites in a city. We show that, after the elicitation, the recommendation problem can be solved as an instance of the non-linear generalized additive knapsack problem.
The enormous number of questions needed to acquire a full preference model when the size of the o... more The enormous number of questions needed to acquire a full preference model when the size of the outcome space is large forces us to work with partial models that approxi- mate the user's preferences. In this way we must devise elicitation strategies that focus on the most important ques- tions and at the same time do not need to enumerate the out- come space. In this paper we focus on adaptive elicitation of GAI-decomposable preferences for top-k recommenda- tion tasks in large combinatorial domains. We propose a method that interleaves the generation of top-k solutions with a heuristic selection of questions for refining the user preference model. Empirical results for a large combinato- rial problem are given.
ABSTRACT We present a new algorithm capable of partitioning sets of objects by taking simultaneou... more ABSTRACT We present a new algorithm capable of partitioning sets of objects by taking simultaneously into account their relational descriptions given by multiple dissimilarity matrices. The novelty of the algorithm is that it is based on a nonlinear aggregation criterion, weighted Tchebycheff distances, more appropriate than linear combinations (such as weighted averages) for the construction of compromise solutions. We obtain a hard partition of the set of objects, the prototype of each cluster and a weight vector that indicates the relevance of each matrix in each cluster. Since this is a clustering algorithm for relational data, it is compatible with any distance function used to measure the dissimilarity between objects. Results obtained in experiments with data sets (synthetic and real) show the usefulness of the proposed algorithm.
2007 2nd International Workshop on Soft Computing Applications, 2007
In this paper we address the multiperson decision making problem when the preferences of each ind... more In this paper we address the multiperson decision making problem when the preferences of each individual is represented by a generalized additive independent (GAI) utility on a product set (the size of which forbids any attempt of exhaustive enumeration of its elements). We focus our attention on finding an optimal compromise solution when it corresponds to maximize a Choquet integral over the elements on the product set. We propose a fast procedure for the exact determination of this optimal compromise solution. This procedure relies on a ranking algorithm that enumerates top-k-solutions of a GAI-Network until a stop condition is met. Finally, we provide results of numerical experiments that indicate the practical efficiency of our procedure.
Feature selection is an important research area that seeks to eliminate unwanted features from da... more Feature selection is an important research area that seeks to eliminate unwanted features from datasets. Many feature selection methods are suggested in the literature, but the evaluation of the best set of features is usually performed using supervised metrics, where labels are required. In this work we propose a methodology that tries to aid data specialists to answer simple but important questions, such as: (1) do current feature selection methods give similar results? (2) is there is a consistently better method ? (3) how to select the m-best features? (4) as the methods are not parameter-free, how to choose the best parameters in the unsupervised scenario? and (5) given different options of selection, could we get better results if we fusion the results of the methods? If yes, how can we combine the results? We analyze these issues and propose a methodology that, based on some unsupervised methods, will make feature selection using strategies that turn the execution of the proc...
ABSTRACT Generalized additive independent (GAI) preference models are, in many situations, more i... more ABSTRACT Generalized additive independent (GAI) preference models are, in many situations, more interesting than additive models, as GAI models allow interdepen-dencies between attributes. However, they are more difficult to construct (elicit), not only because the number of questions needed to be asked increases, but also because we must know what the interdependencies are, i.e. the structure of the model. We introduce the use of association rules to select attributes and detect simple interactions between them during an interactive preference elicitation pro-cess that intends to build a GAI utility function for the preferences of the user of a recommender system. Using this strategy, we have built a recommender system pro-totype that suggests touristic sites in a city. We show that, after the elicitation, the recommendation problem can be solved as an instance of the non-linear generalized additive knapsack problem.
The enormous number of questions needed to acquire a full preference model when the size of the o... more The enormous number of questions needed to acquire a full preference model when the size of the outcome space is large forces us to work with partial models that approxi- mate the user's preferences. In this way we must devise elicitation strategies that focus on the most important ques- tions and at the same time do not need to enumerate the out- come space. In this paper we focus on adaptive elicitation of GAI-decomposable preferences for top-k recommenda- tion tasks in large combinatorial domains. We propose a method that interleaves the generation of top-k solutions with a heuristic selection of questions for refining the user preference model. Empirical results for a large combinato- rial problem are given.
ABSTRACT We present a new algorithm capable of partitioning sets of objects by taking simultaneou... more ABSTRACT We present a new algorithm capable of partitioning sets of objects by taking simultaneously into account their relational descriptions given by multiple dissimilarity matrices. The novelty of the algorithm is that it is based on a nonlinear aggregation criterion, weighted Tchebycheff distances, more appropriate than linear combinations (such as weighted averages) for the construction of compromise solutions. We obtain a hard partition of the set of objects, the prototype of each cluster and a weight vector that indicates the relevance of each matrix in each cluster. Since this is a clustering algorithm for relational data, it is compatible with any distance function used to measure the dissimilarity between objects. Results obtained in experiments with data sets (synthetic and real) show the usefulness of the proposed algorithm.
2007 2nd International Workshop on Soft Computing Applications, 2007
In this paper we address the multiperson decision making problem when the preferences of each ind... more In this paper we address the multiperson decision making problem when the preferences of each individual is represented by a generalized additive independent (GAI) utility on a product set (the size of which forbids any attempt of exhaustive enumeration of its elements). We focus our attention on finding an optimal compromise solution when it corresponds to maximize a Choquet integral over the elements on the product set. We propose a fast procedure for the exact determination of this optimal compromise solution. This procedure relies on a ranking algorithm that enumerates top-k-solutions of a GAI-Network until a stop condition is met. Finally, we provide results of numerical experiments that indicate the practical efficiency of our procedure.
Uploads