notes.txt

Four points are:
	estimate the size of development
	estimate the effort of the person
	estimate the time in months or years
	estimate the cost in dollars

there is no specific method to determine the cost estimate

prediction are classified into three
	expert judgement (can be used as an estimte to algo models)
	algorithmic models (main criterion is size of source file)
		cocomo model
		function point anaylsis
		slim model
	machine learning

ann gave more promising results than regression techniques
MLPNN and a stepwise regression

NN gives some good results compared to linear regression

ann in software focused on accuracy comparsion rather than suitability

factors like configuarbility and exploratory are also equally important

linear regression is sensitive to outliers

cost drivers are like categorical variables (each of it has a specific value which is called as effort multiplier)

datasets were divided in the range of every sixth project from the training data.

To apply the linear regression, the data needs to be normally distributed.
Tranform it using a log to base e conversion to reduce the variance.
GRM(General Regression Models), apply adjusted-R2 rather than R2