Predicting Software Development Effort Using Tuned Artificial Neural Network

Abstract

Software development effort prediction using fixed mathematical formulae is inadequate due to impreciseness and nonlinearity exist in the software project data and leads to high prediction error rate, on the other hand Artificial Neural Network (ANN) techniques are very popular for prediction of software development effort due to its capability to map non linear input with output. This paper explores Error Back Propagation Network (EBPN) for software development effort prediction, by tuning some algorithm specific parameters like learning rate and momentum. EBPN is trained with two benchmark data sets: China and Maxwell. Results are analyzed in terms of various measures and found to be satisfactory.

The application of artificial intelligence in software engineering

“…Hota et. al, [49] explored Error Back Propagation Network (EBPN) for software development effort prediction, by tuning two algorithm specific parameters, learning rate and momentum. EBPN is a kind of NN popularly used as predictor due to its capability of mapping high dimensional data.…”

Level of Confidence in Software Effort Estimation by an Intelligent Fuzzy – Neuro - Genetic Approach

“…The preeminent efficiency of genetically trained neural networks is mentioned in various researches. For instance After a series of experiments and simulations Shukla [22] concluded that the genetically trained neural networks outperforms back propagation trained and quick propagation trained neural networks in software effort estimation. The recommended model evaluates software development effort in the function of seventeen cost drivers and five scale factors.…”

“…A convenient way is to try many different networks, calculate the generalization error for each network and select the network with minimum generalization error [23].Following the above rule, we performed a number of hit and trial experiments from 2 to 20 nodes. [22] concluded that the genetically trained neural networks outperforms back propagation trained and quick propagation trained neural networks in software effort estimation The suitable values for control parameters of genetic algorithm have been found by running various simulations and have been listed in Table 1. We have used binary string chromosomes.…”

A differential evolution-based model to estimate the software services development effort

“…Different kinds of regression , COCOMO models and COCOMO II are the most famous algorithmic models, and ABE , CART , expert judgment , artificial neural network , learning and artificial intelligence techniques , fuzzy rules, and optimization algorithms are the most popular non‐algorithmic methods. ABE method uses the two similarity and solution functions for effort estimation.…”