Navegando por Palavras-chave "Kaizen Programming"

Agora exibindo 1 - 4 de 4
  • Nenhuma Miniatura disponível
    Item
    Somente Metadadados
    Classification of cardiac arrhythmia by random forests with features constructed by kaizen programming with linear genetic programming
    (Wiley, 2016) Sotto, Leo F. D. P. [UNIFESP]; Coelho, Regina C. [UNIFESP]; de Melo, Vinicius V. [UNIFESP]
    Cardiac rhythm disorders may cause severe heart diseases, stroke, and even sudden cardiac death. Some arrhythmias are so serious that can cause injury to other organs, for instance, brain, kidneys, lungs or liver. Therefore, early and correct diagnosis of cardiac arrhythmia is essential to the prevention of serious problems. There are expert systems to classify arrhythmias from electrocardiograms signals. However, it has been shown that not only selecting the correct features from the dataset but also generating combined features could be the key to having real progress in classification. Therefore, this paper investigates a novel hybrid evolutionary technique to perform both tasks at the same time, finding complementary features that cover different characteristics of the data. The new features were tested with a widely-used classifier called Random Forests. The method reduced a dataset with 279 attributes to 26 attributes and achieved accuracies of 86.39% for binary classification and 77.69% for multiclass. Our approach outperformed several popular feature selection, feature generation, and state-of-the-art related work from the literature.
  • Nenhuma Miniatura disponível
    Item
    Somente Metadadados
    Improving logistic regression classification of credit approval with features constructed by kaizen programming
    (Funpec-Editora, 2016) de Melo, Vinicius Veloso [UNIFESP]; Banzhaf, Wolfgang
    In this contribution, we employ the recently proposed Kaizen Programming (KP) approach to fi nd high-quality nonlinear combinations of the original features in a dataset. KP constructs many complementary features at the same time, which are selected by their importance, not by model quality. We investigated our approach in a well-known realworld credit scoring dataset. When compared to related approaches, KP reaches similar or better results, but evaluates fewer models.
  • Nenhuma Miniatura disponível
    Item
    Somente Metadadados
    Improving the prediction of material properties of concrete using Kaizen Programming with Simulated Annealing
    (Elsevier Science Bv, 2017) de Melo, Vinicius Veloso [UNIFESP]; Banzhaf, Wolfgang
    Predicting the properties of materials like concrete has been proven a difficult task given the complex interactions among its components. Over the years, researchers have used Statistics, Machine Learning, and Evolutionary Computation to build models in an attempt to accurately predict such properties. High quality models are often non-linear, justifying the study of nonlinear regression tools. In this paper, we employ a traditional multiple linear regression method by ordinary least squares to solve the task. However, the model is built upon nonlinear features automatically engineered by Kaizen Programming, a recently proposed hybrid method. Experimental results show that Kaizen Programming can find low correlated features in an acceptable computational time. Such features build high-quality models with better predictive quality than results reported in the literature. (C) 2017 Elsevier B.V. All rights reserved.
  • Nenhuma Miniatura disponível
    Item
    Somente Metadadados
    Solving the lawn mower problem with kaizen programming and lambda-linear genetic programming for module acquisition
    (Associacao Paulista Medicina, 2016) dal Piccol Sotto, Leo Francoso [UNIFESP]; de Melo, Vinicius Veloso [UNIFESP]
    In this work, we have tested a new approach for evolving modular programs: Kaizen Programming (KP) with lambda-Linear Genetic Programming (lambda-LGP) and a heuristic search procedure to solve the well-known Lawn Mower problem. KP is a novel hybrid approach that tries to efficiently combine partial solutions to generate a high-quality complete solution. Being a hybrid, KP may use different types of methods to generate partial solutions, assess their importance to the complete solution, and solve the complete problem. Experiments on the Lawn Mower problem show that the proposed method is effective in finding the expected solution. It is a new alternative for evolving modular programs, but further investigations are necessary to improve its performance.