On the Problem of Attribute Selection for Software Cost Estimation: Input Backward Elimination Using Artificial Neural Networks

Papatheocharous, Efi; Andreou, Andreas S.

doi:10.1007/978-3-642-16239-8_38

Cited by 6 publications

(4 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The sequential backward selection technique removes one least important feature from the subset, based on a fixed criterion in every iteration. The selection process starts with a full set of features and stops when a predefined threshold for the number of variables is reached or further removal of features does not improve the performance [12]. We used mutual information as the selection criteria and set the selection threshold to seven variables.…”

Section: Methodsmentioning

confidence: 99%

Feature selection and transformation by machine learning reduce variable numbers and improve prediction for heart failure readmission or death

et al. 2019

View full text Add to dashboard Cite

Background The prediction of readmission or death after a hospital discharge for heart failure (HF) remains a major challenge. Modern healthcare systems, electronic health records, and machine learning (ML) techniques allow us to mine data to select the most significant variables (allowing for reduction in the number of variables) without compromising the performance of models used for prediction of readmission and death. Moreover, ML methods based on transformation of variables may potentially further improve the performance. Objective To use ML techniques to determine the most relevant and also transform variables for the prediction of 30-day readmission or death in HF patients. Methods We identified all Western Australian patients aged 65 years and above admitted for HF between 2003–2008 in linked administrative data. We evaluated variables associated with HF readmission or death using standard statistical and ML based selection techniques. We also tested the new variables produced by transformation of the original variables. We developed multi-layer perceptron prediction models and compared their predictive performance using metrics such as Area Under the receiver operating characteristic Curve (AUC), sensitivity and specificity. Results Following hospital discharge, the proportion of 30-day readmissions or death was 23.7% in our cohort of 10,757 HF patients. The prediction model developed by us using a smaller set of variables ( n = 8) had comparable performance (AUC 0.62) to the traditional model ( n = 47, AUC 0.62). Transformation of the original 47 variables further improved (p<0.001) the performance of the predictive model (AUC 0.66). Conclusions A small set of variables selected using ML matched the performance of the model that used the full set of 47 variables for predicting 30-day readmission or death in HF patients. Model performance can be further significantly improved by transforming the original variables using ML methods.

show abstract

Section: Methodsmentioning

confidence: 99%

Feature selection and transformation by machine learning reduce variable numbers and improve prediction for heart failure readmission or death

et al. 2019

View full text Add to dashboard Cite

show abstract

“…In the second procedure, the neural network model is estimated with all the selected variables as inputs, and the relative importance of every input variable is calculated. As a nonlinear model, determining the relative importance of variables is more difficult than in linear regression models, and Garson's method is used in this study, because it was proved that the neural network can identify the most influential input variables from a given variable list by Garson's method (Garson 1991), and recently, the reliability of Garson's method in the variable selection process of three-layer neural network is verified to be better than other widely used methods, such as correlation method and principal component analysis (Papatheocharous and Andreou 2010;Fischer 2015;Yousefi et al 2018;Liu et al 2018). According to Garson's method (Garson 1991), for a neural network model with N neurons in the input layer and L neurons in the hidden layer, the relative importance of the ith input variable to the kth output variable ( I ik ) can be defined as where ij is the weight of the ith neuron in the input layer and jth neuron in the hidden layer, and jk is the weight of the jth neuron in the hidden layer and kth neuron in the output layer.…”

Section: Neural Network Model Selection Processmentioning

confidence: 99%

“…After determining the input variables, output variables, and number of neurons in the hidden layer, the actual structure of the neural network model is determined. Garson's method is capable of quantifying the relative importance and select variables (Papatheocharous and Andreou 2010;Fischer 2015;Yousefi et al 2018;Liu et al 2018). Other methods such as Olden's method and SHapley Additive exPlanations (SHAP) values can quantify the intensity and direction of each input variable's contribution to each output variable (Olden and Jackson 2002;Lundberg and Lee 2017), which is helpful to understand the relationship between each input and output variable in the model.…”

Section: Parameter Analysismentioning

confidence: 99%

Neural network model for predicting variation in walking dynamics of pedestrians in social groups

Sun

Duives

et al. 2022

Transportation

View full text Add to dashboard Cite

show abstract

“…Functional Link Artificial Neural Network (FLANN) based software cost estimation (FBE), which is essentially a machine learning technique, was introduced by Rao et al (2009). Due to its conceptual simplicity and empirical competitiveness, FLANN has been extensively studied and applied (Jodpimai et al, 2010;Papatheocharous et al, 2010;Abhishek et al, 2010;Khatibi et al, 2011). FLANN is basically a flat net and the need of the hidden layer is removed and hence, the BP learning algorithm used in this network becomes very simple, originally proposed by Pao et al (1992).…”

Section: Introductionmentioning

confidence: 99%

Functional Link Artificial Neural Networks for Software Cost Estimation

Dehuri¹,

Rao

Mall

2012

International Journal of Applied Evolutionary Computation

View full text Add to dashboard Cite

Software cost estimation is the process of predicting the effort required to develop a software system. Software development projects often overrun their planned effort as defined at preliminary design review. Software cost estimation is important for budgeting, risk analysis, project planning, and software improvement analysis. In this paper, the authors propose a faster functional link artificial neural network (FLANN) based software cost estimation. By means of preprocessing, i.e., optimal reduced datasets (ORD), the authors make the functional link artificial neural network faster. Optimal reduced datasets, which reduce the whole project base into small subsets that consist of only representative projects. The representative projects are given as input to FLANN and tested on eight state-of-the-art polynomial expansions. The proposed methods are validated on five real time datasets. This approach yields accurate results vis-à-vis conventional FLANN, support vector machine regression (SVR), radial basis function (RBF), classification, and regression trees (CART).

show abstract

On the Problem of Attribute Selection for Software Cost Estimation: Input Backward Elimination Using Artificial Neural Networks

Cited by 6 publications

References 10 publications

Feature selection and transformation by machine learning reduce variable numbers and improve prediction for heart failure readmission or death

Feature selection and transformation by machine learning reduce variable numbers and improve prediction for heart failure readmission or death

Neural network model for predicting variation in walking dynamics of pedestrians in social groups

Functional Link Artificial Neural Networks for Software Cost Estimation

Contact Info

Product

Resources

About