Fig. 2From: A genetic algorithm based framework for software effort predictionExperimental Design Steps: This figure represents the experimental design steps. In this experiment, the set N−P A S S=10 and computed the Correlation Coefficient (CC) average is set after N−P A S S runs. For each PASS, 90% of the data as historical at random is selected. An N=10 x M=10-fold cross-validation is used to evaluate each learning scheme. The evaluation metrics were computed after N x M-fold cross-validation. The fitness function of each genetic individual was executed in 1000 holdout experiments, (N−P A S S=10) and N=10 x M=10-fold cross-validation. The mean of the 1000 CC measures was reported as the evaluation performance per genetic individual. The historicalData (90%) was preprocessed considering preprocessing techniques. Then, the predictor was used to predict with the newData (10%), which was preprocessed the same way as the historical dataBack to article page