Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00045
S. Chevtchenko, Rafaella F. Vale, F. Cordeiro, V. Macário
Convolutional architectures have in recent years become state-of-the-art for several object detection tasks. However, these detectors have not yet been evaluated for detection and monitoring of beach areas. As some of these areas need to be continually monitored for dangerous situations, such as shark attacks, an automated system would be an effective risk control measure. The most significant and specific challenges for this problem are variable scene illumination, partial occlusion and distant camera position. In this work we present a study on three recent convolutional architectures for the task of people detection in beach scenarios. Our dataset is composed of images taken in the Boa Viagem beach, in Brazil, and is used to evaluate Faster R-CNN, R-FCN and SSD in terms of quality and speed of detection. The detectors are pretrained on a dataset containing 91 classes of objects, including people with different levels of scale and occlusion. The results suggest that the Faster R-CNN meta-architecture with the Resnet 101 feature extractor generates significantly better detections in terms of F-measure, while performing at 5.6 fps on a GTX 1080 Ti GPU.
卷积架构近年来已经成为一些目标检测任务的最先进的技术。然而，这些探测器在探测和监测海滩地区方面尚未得到评价。由于其中一些区域需要持续监测危险情况，例如鲨鱼袭击，自动化系统将是一种有效的风险控制措施。这个问题最重要和最具体的挑战是可变的场景照明，部分遮挡和远距离摄像机位置。在这项工作中，我们对海滩场景中人员检测任务的三种最新卷积架构进行了研究。我们的数据集由在巴西Boa Viagem海滩拍摄的图像组成，并用于评估Faster R-CNN, R-FCN和SSD在检测质量和速度方面的性能。检测器在包含91类对象的数据集上进行预训练，包括具有不同规模和遮挡水平的人。结果表明，更快的R-CNN元架构与Resnet 101特征提取器在F-measure方面产生了显着更好的检测，而在GTX 1080 Ti GPU上以5.6 fps的速度执行。
{"title":"Deep Learning for People Detection on Beach Images","authors":"S. Chevtchenko, Rafaella F. Vale, F. Cordeiro, V. Macário","doi":"10.1109/BRACIS.2018.00045","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00045","url":null,"abstract":"Convolutional architectures have in recent years become state-of-the-art for several object detection tasks. However, these detectors have not yet been evaluated for detection and monitoring of beach areas. As some of these areas need to be continually monitored for dangerous situations, such as shark attacks, an automated system would be an effective risk control measure. The most significant and specific challenges for this problem are variable scene illumination, partial occlusion and distant camera position. In this work we present a study on three recent convolutional architectures for the task of people detection in beach scenarios. Our dataset is composed of images taken in the Boa Viagem beach, in Brazil, and is used to evaluate Faster R-CNN, R-FCN and SSD in terms of quality and speed of detection. The detectors are pretrained on a dataset containing 91 classes of objects, including people with different levels of scale and occlusion. The results suggest that the Faster R-CNN meta-architecture with the Resnet 101 feature extractor generates significantly better detections in terms of F-measure, while performing at 5.6 fps on a GTX 1080 Ti GPU.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128039211","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00061
T. Filisbino, G. Giraldi, C. Thomaz
In this paper we present a nonlinear version of the discriminant principal component analysis, named NDPCA, that is based on kernel support vector machines (KSVM) and the AdaBoost technique. Specifically, the problem of ranking principal components, computed from two-class databases, is addressed by applying the AdaBoost procedure in a nested loop: each iteration of the inner loop boosts weak classifiers to a moderate one while the outer loop combines the moderate classifiers to build the global discriminant vector. In the proposed NDPCA, each weak learner is a linear classifier computed through a separating hyperplane defined by a KSVM decision boundary in the PCA space. We compare the proposed methodology with counterpart ones using facial expressions of the Radboud and Jaffe image databases. Our experimental results have shown that NDPCA outperforms the PCA in classification tasks. Also, it is competitive if compared with counterpart techniques given also suitable results for reconstruction.
{"title":"Nonlinear Discriminant Principal Component Analysis for Image Classification and Reconstruction","authors":"T. Filisbino, G. Giraldi, C. Thomaz","doi":"10.1109/BRACIS.2018.00061","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00061","url":null,"abstract":"In this paper we present a nonlinear version of the discriminant principal component analysis, named NDPCA, that is based on kernel support vector machines (KSVM) and the AdaBoost technique. Specifically, the problem of ranking principal components, computed from two-class databases, is addressed by applying the AdaBoost procedure in a nested loop: each iteration of the inner loop boosts weak classifiers to a moderate one while the outer loop combines the moderate classifiers to build the global discriminant vector. In the proposed NDPCA, each weak learner is a linear classifier computed through a separating hyperplane defined by a KSVM decision boundary in the PCA space. We compare the proposed methodology with counterpart ones using facial expressions of the Radboud and Jaffe image databases. Our experimental results have shown that NDPCA outperforms the PCA in classification tasks. Also, it is competitive if compared with counterpart techniques given also suitable results for reconstruction.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"70 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131067809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00068
G. Fritsche, A. Pozo
Many-objective optimization problems (MaOPs) are a great challenge for multi-objective evolutionary algorithms (MOEAs) and lately, several MOEAs have been proposed. Each MOEA uses different algorithmic components during the search process and performs differently. Therefore, there is no single algorithm able to achieve the best results in all problems. The collaboration of multiple MOEAs and the use of hyperheuristics can help to create a searchability able to achieve good results in a wide range of problem instances. In this context, this research proposes a model for collaboration of MOEAs guided by hyper-heuristic, called HHcMOEA. In HHcMOEA, the hyper-heuristic controls and mix MOEAs, automatically deciding which one to apply during the search process. On the other hand, HHcMOEA also incorporates exchange of information between the MOEAs. And, a fitness improvement rate metric, based on the R2 indicator to decide about the quality of the application of an MOEA. HHcMOEA is implemented using a set of MOEAs with diverse characteristics. An experiment is used to evaluate HHcMOEA in two versions: with and without information exchange. Although, the two versions of HHcMOEA are compared to the MOEAs applied alone. The empirical evaluation used a set of benchmark problems with different properties. The proposed model achieved the best result or equivalent to the best in almost all problems. Still, the results were deteriorated when the information exchange strategy was not used.
{"title":"A Hyper-Heuristic Collaborative Multi-objective Evolutionary Algorithm","authors":"G. Fritsche, A. Pozo","doi":"10.1109/BRACIS.2018.00068","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00068","url":null,"abstract":"Many-objective optimization problems (MaOPs) are a great challenge for multi-objective evolutionary algorithms (MOEAs) and lately, several MOEAs have been proposed. Each MOEA uses different algorithmic components during the search process and performs differently. Therefore, there is no single algorithm able to achieve the best results in all problems. The collaboration of multiple MOEAs and the use of hyperheuristics can help to create a searchability able to achieve good results in a wide range of problem instances. In this context, this research proposes a model for collaboration of MOEAs guided by hyper-heuristic, called HHcMOEA. In HHcMOEA, the hyper-heuristic controls and mix MOEAs, automatically deciding which one to apply during the search process. On the other hand, HHcMOEA also incorporates exchange of information between the MOEAs. And, a fitness improvement rate metric, based on the R2 indicator to decide about the quality of the application of an MOEA. HHcMOEA is implemented using a set of MOEAs with diverse characteristics. An experiment is used to evaluate HHcMOEA in two versions: with and without information exchange. Although, the two versions of HHcMOEA are compared to the MOEAs applied alone. The empirical evaluation used a set of benchmark problems with different properties. The proposed model achieved the best result or equivalent to the best in almost all problems. Still, the results were deteriorated when the information exchange strategy was not used.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128364364","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/bracis.2018.00075
Saulo Martiello Mastelini, Everton José Santana, Victor Guilherme Turrisi da Costa, Sylvio Barbon Junior
Machine learning methods for multi-target regression (MTR) rely on the hypothesis that an inter-target correlation can improve predictive performance. In the last years, many MTR methods were developed, but there are still questions about how their performances are influenced by the datasets characteristics such as linearity, number of targets, and inter-correlation complexity. Aiming at contributing to the understanding of the relationship between the dataset properties and MTR methods, we generated 33 synthetic datasets with controlled characteristics and tested their performance with single-target and six MTR methods. The results showed that MTR methods were able to improve performance even in datasets whose targets were not linearly correlated among them, but the predictive improvement differed among the combinations of method/regressor according to the dataset composition.
{"title":"Benchmarking Multi-target Regression Methods","authors":"Saulo Martiello Mastelini, Everton José Santana, Victor Guilherme Turrisi da Costa, Sylvio Barbon Junior","doi":"10.1109/bracis.2018.00075","DOIUrl":"https://doi.org/10.1109/bracis.2018.00075","url":null,"abstract":"Machine learning methods for multi-target regression (MTR) rely on the hypothesis that an inter-target correlation can improve predictive performance. In the last years, many MTR methods were developed, but there are still questions about how their performances are influenced by the datasets characteristics such as linearity, number of targets, and inter-correlation complexity. Aiming at contributing to the understanding of the relationship between the dataset properties and MTR methods, we generated 33 synthetic datasets with controlled characteristics and tested their performance with single-target and six MTR methods. The results showed that MTR methods were able to improve performance even in datasets whose targets were not linearly correlated among them, but the predictive improvement differed among the combinations of method/regressor according to the dataset composition.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"98 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122760121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/bracis.2018.00016
Igor Cataneo Silveira, Denis Deratani Mauá
Answering questions formulated in natural language is a long standing quest in Artificial Intelligence. However, even formulating the problem in precise terms has proven to be too challenging, which lead many researchers to focus on Multiple-Choice Question Answering problems. One particularly interesting type of the latter problem is solving standardized tests such as university entrance exams. The Exame Nacional do Ensino Médio (ENEM) is a High School level exam widely used by Brazilian universities as entrance exam, and the world's second biggest university entrance examination in number of registered candidates. In this work we tackle the problem of answering purely textual multiple-choice questions from the ENEM. We build on a previous solution that formulated the problem as a text information retrieval problem. In particular, we investigate how to enhance these methods by text augmentation using Word Embedding and WordNet, a structured lexical database where words are connected according to some relations like synonymy and hypernymy. We also investigate how to boost performance by building ensembles of weakly correlated solvers. Our approaches obtain accuracies ranging from 26% to 29.3%, outperforming the previous approach.
{"title":"Advances in Automatically Solving the ENEM","authors":"Igor Cataneo Silveira, Denis Deratani Mauá","doi":"10.1109/bracis.2018.00016","DOIUrl":"https://doi.org/10.1109/bracis.2018.00016","url":null,"abstract":"Answering questions formulated in natural language is a long standing quest in Artificial Intelligence. However, even formulating the problem in precise terms has proven to be too challenging, which lead many researchers to focus on Multiple-Choice Question Answering problems. One particularly interesting type of the latter problem is solving standardized tests such as university entrance exams. The Exame Nacional do Ensino Médio (ENEM) is a High School level exam widely used by Brazilian universities as entrance exam, and the world's second biggest university entrance examination in number of registered candidates. In this work we tackle the problem of answering purely textual multiple-choice questions from the ENEM. We build on a previous solution that formulated the problem as a text information retrieval problem. In particular, we investigate how to enhance these methods by text augmentation using Word Embedding and WordNet, a structured lexical database where words are connected according to some relations like synonymy and hypernymy. We also investigate how to boost performance by building ensembles of weakly correlated solvers. Our approaches obtain accuracies ranging from 26% to 29.3%, outperforming the previous approach.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121234129","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00098
T. Rocha, Ana Teresa C. Martins, F. Ferreira
We define a propositional substring logic (PS) in which atomic sentences represent substring properties of strings. We also investigate the following variation of the boolean function synthesis (BFS) problem: given a sample of classified strings, find a PS formula in disjunctive normal form with the minimum number of clauses and consistent with the sample. We call this problem PS formula synthesis (PSFS). The advantages of using PS is that it is as expressive as first-order logic (FO) over strings with the successor relation, and PS formulas are more succinct than FO formulas. We show that PSFS is NP-complete, and we propose an algorithm to solve PSFS via a reduction to the BFS problem.
{"title":"Synthesis of a DNF Formula From a Sample of Strings","authors":"T. Rocha, Ana Teresa C. Martins, F. Ferreira","doi":"10.1109/BRACIS.2018.00098","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00098","url":null,"abstract":"We define a propositional substring logic (PS) in which atomic sentences represent substring properties of strings. We also investigate the following variation of the boolean function synthesis (BFS) problem: given a sample of classified strings, find a PS formula in disjunctive normal form with the minimum number of clauses and consistent with the sample. We call this problem PS formula synthesis (PSFS). The advantages of using PS is that it is as expressive as first-order logic (FO) over strings with the successor relation, and PS formulas are more succinct than FO formulas. We show that PSFS is NP-complete, and we propose an algorithm to solve PSFS via a reduction to the BFS problem.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126554645","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00026
Miguel D. de S. Wanderley, R. Prudêncio
Convolutional Neural Networks have been successfully applied in several image related tasks. On another hand, there are some overhead costs in most of the real applications. Often, the Deep Learning techniques demand a huge amount of data for training and also a crescent need for handling high definition images. For this reason, late network architectures are getting even more complex and deeper. These factors lead to a long training time even when specific hardware is available. In this paper, we present a novel incremental training procedure which is able to train faster with small performance losses, based on measuring and ordering the relative complexity of subsets of the training set. The findings reveal an expressive reduction in the number of training steps, without critical performance losses. Experiments showed that the proposed method can be about 40% faster, with less than 10% of accuracy loss.
{"title":"Increasing Convolutional Neural Networks Training Speed by Incremental Complexity Learning","authors":"Miguel D. de S. Wanderley, R. Prudêncio","doi":"10.1109/BRACIS.2018.00026","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00026","url":null,"abstract":"Convolutional Neural Networks have been successfully applied in several image related tasks. On another hand, there are some overhead costs in most of the real applications. Often, the Deep Learning techniques demand a huge amount of data for training and also a crescent need for handling high definition images. For this reason, late network architectures are getting even more complex and deeper. These factors lead to a long training time even when specific hardware is available. In this paper, we present a novel incremental training procedure which is able to train faster with small performance losses, based on measuring and ordering the relative complexity of subsets of the training set. The findings reveal an expressive reduction in the number of training steps, without critical performance losses. Experiments showed that the proposed method can be about 40% faster, with less than 10% of accuracy loss.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124677530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00081
Rodrigo Caputo, Edmilson Santos
Since 1997, RoboCup organizes robotics competitions in order to disseminate and promote technological advancement worldwide. One of the platforms created by the RoboCup is robot soccer, which consists of disputes between two different teams of autonomous robotic agents who play soccer according to pre-established rules. In this scenario, several researches have already been conducted to find an efficient strategy to manage the players in a totally autonomous way. This paper presents a method based on a hierarchical control system called STP (Skills, Tactics and Plays) for decision making in robot soccer within the Small Size category. Our main goal is to apply Bayesian classifiers supported by ranking for choosing the appropriate Play (from STP model) to be performed according to the state of the game. We have evaluated the results of three Bayesian classifiers: Naive Bayes, TAN and K2. Empirical results obtained in the initial experiments indicate that the proposed method is promising, and it tends to be tolerant to classification errors.
自1997年以来，RoboCup组织机器人比赛，以传播和促进全球技术进步。机器人世界杯创造的平台之一是机器人足球，它由两支不同的自主机器人代理球队根据预先制定的规则进行足球比赛。在这种情况下，已经进行了一些研究，以找到一种有效的策略，以完全自主的方式管理玩家。本文提出了一种基于分层控制系统STP (Skills, Tactics and Plays)的小型机器人足球决策方法。我们的主要目标是应用由排名支持的贝叶斯分类器，根据游戏的状态选择合适的Play(来自STP模型)来执行。我们评估了三种贝叶斯分类器的结果:朴素贝叶斯，TAN和K2。初步实验结果表明，该方法具有较强的分类容错性，具有较好的应用前景。
{"title":"Bayesian Classifiers Supported by Ranking for Decision Making in Robot Soccer","authors":"Rodrigo Caputo, Edmilson Santos","doi":"10.1109/BRACIS.2018.00081","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00081","url":null,"abstract":"Since 1997, RoboCup organizes robotics competitions in order to disseminate and promote technological advancement worldwide. One of the platforms created by the RoboCup is robot soccer, which consists of disputes between two different teams of autonomous robotic agents who play soccer according to pre-established rules. In this scenario, several researches have already been conducted to find an efficient strategy to manage the players in a totally autonomous way. This paper presents a method based on a hierarchical control system called STP (Skills, Tactics and Plays) for decision making in robot soccer within the Small Size category. Our main goal is to apply Bayesian classifiers supported by ranking for choosing the appropriate Play (from STP model) to be performed according to the state of the game. We have evaluated the results of three Bayesian classifiers: Naive Bayes, TAN and K2. Empirical results obtained in the initial experiments indicate that the proposed method is promising, and it tends to be tolerant to classification errors.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122188000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/bracis.2018.00071
Everton Rodrigues Reis, Jaime Simão Sichman
Portfolio management is a challenging task where humans have to make decisions under uncertainty. Since usually humans tend to avoid unknown risk, in general they don't maximize their utility function when managing a portfolio. This fact favours using an automated trading system for portfolio management. In this work, we propose an automated trading system using multiagent systems. We use fundamental and cluster analysis to select the stocks, and additionally we employ a financial distress prediction model to estimate companies financial health. We also optimize the portfolio for different investor's utility functions. Comparing our approach's results to a benchmark, we have obtained higher return values and lower risks; moreover, the approach was profitable even when we have added brokerage fees.
{"title":"MAVIS: A Multiagent Value Investing System","authors":"Everton Rodrigues Reis, Jaime Simão Sichman","doi":"10.1109/bracis.2018.00071","DOIUrl":"https://doi.org/10.1109/bracis.2018.00071","url":null,"abstract":"Portfolio management is a challenging task where humans have to make decisions under uncertainty. Since usually humans tend to avoid unknown risk, in general they don't maximize their utility function when managing a portfolio. This fact favours using an automated trading system for portfolio management. In this work, we propose an automated trading system using multiagent systems. We use fundamental and cluster analysis to select the stocks, and additionally we employ a financial distress prediction model to estimate companies financial health. We also optimize the portfolio for different investor's utility functions. Comparing our approach's results to a benchmark, we have obtained higher return values and lower risks; moreover, the approach was profitable even when we have added brokerage fees.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125711293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-10-01DOI: 10.1109/BRACIS.2018.00013
Vitor de Albuquerque Torreao, Renato Vimieiro
Many Evolutionary Algorithms have been proposed to solve the Subgroup Discovery task. Some of these, however, have been shown to work poorly in high dimensional problems. The best performing evolutionary algorithm for subgroup discovery in high dimensional datasets has a particular way to initialize its starting population, limiting the size of initial solutions to the lowest possible value. As with most population-based techniques, the outcome of evolutionary algorithms is usually dependent on the initial set of solutions, which are typically randomly generated. The impact of choosing one initialization technique over another in the final presented solution has been the topic of many published works in the broad area of evolutionary computation. However, to the best of our knowledge, it has not been the topic of study in the specific case of the Subgroup Discovery task, especially when considering high dimensional datasets. Therefore, this paper aims at studying whether or not it is possible to improve the performance of evolutionary algorithms in high dimensional subgroup discovery tasks by biasing the initial population to individuals with lower sizes.
{"title":"Effects of Population Initialization on Evolutionary Techniques for Subgroup Discovery in High Dimensional Datasets","authors":"Vitor de Albuquerque Torreao, Renato Vimieiro","doi":"10.1109/BRACIS.2018.00013","DOIUrl":"https://doi.org/10.1109/BRACIS.2018.00013","url":null,"abstract":"Many Evolutionary Algorithms have been proposed to solve the Subgroup Discovery task. Some of these, however, have been shown to work poorly in high dimensional problems. The best performing evolutionary algorithm for subgroup discovery in high dimensional datasets has a particular way to initialize its starting population, limiting the size of initial solutions to the lowest possible value. As with most population-based techniques, the outcome of evolutionary algorithms is usually dependent on the initial set of solutions, which are typically randomly generated. The impact of choosing one initialization technique over another in the final presented solution has been the topic of many published works in the broad area of evolutionary computation. However, to the best of our knowledge, it has not been the topic of study in the specific case of the Subgroup Discovery task, especially when considering high dimensional datasets. Therefore, this paper aims at studying whether or not it is possible to improve the performance of evolutionary algorithms in high dimensional subgroup discovery tasks by biasing the initial population to individuals with lower sizes.","PeriodicalId":405190,"journal":{"name":"2018 7th Brazilian Conference on Intelligent Systems (BRACIS)","volume":" 23","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133121923","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：