Jan. 31, 2014
cs562

Review of Effort/Time estimation

Characterization
Tracking
Evaluation
Prediction
Improvement

Hard to come up with a good model

Planning poker is easy but there are lots of other methods

Conceptualize entity of interest

GQM - goal question metric model
Want conceptual, operational, quantitative level goals
	-Idea: How can I make my code run faster
	-Questions related to goal: How will it affect debugging time
	-Determine metric for goal

Software Engineering Time Estimation (project management)

	Divide and Conquer method
	Include everything
	Use past experience

	Analysis Effort Method
		Estimate duration of projects
		Good for producing initial estimates
		Inputs are numeric factors
			Size (S)
			Familiarity (F)
			Complexity©
		Use with the duration of preparing specification to find length of the following phases
		within a lookup table
			Design
			Coding
			Testing (including unit testing)
		Doesn't include training times
	COCOMO (constructive cost model) by Barry Boehm (1981)
		Cost estimation model
		Don't use for real projects

		Applies to three different classes of projects
			"Organic" - small teams with experience working with loose requirements
			"Semi-detached projects" - "medium" teams with mixed experience with mixed 			requirements
			"Embedded projects" - tight constraints, may be combination of semi-detached
			projects

		Compute SW dev effort as a function of program size

		Effort Applied = a_b * KLOC^b_b  (man months)
		Development Time = c_b * Effort_Applied ^ d_b (months)
		People Required =  Effort_Applied / Development_Time (num people)

		Constants defined using a table

		Organic - 2.4  1.05  2.5  0.38
		Semi-detached - 3.0  1.12  2.5  0.35
		Embedded - 3.6  1.2  2.5  0.32

		Intermediate COCOMO
			Use 15 attributes rated from 1 to 6 as effort adjustment factors corresponding
			to varying values between 0.70 and 1.65
			Includes things like 
				SW reliability
				Size of database
				Complexity of project
				HW attributes
				Personnel attributes
				Project attributes
				etc.

			New effort formula is
			E = a_i * KLoC^b_i for each Factor above
		
		http://en.wikipedia.org/wiki/COCOMO
		http://www.codeproject.com/Articles/9266/Software-Project-Cost-Estimates-Using-COCOMO-II-Mo

	Function point analysis
		Analyze functionality rather than LoC
		Used for many years (since 1983)
		Sum of Features * Weights using industry standards (lookup table)

		Consider:
			External inputs
			External outputs
			External inquiries
			Internal logical files
			External interface files

		Components are assigned weights based upon simple descriptions of the project
			Simple
			Average
			Complex

		Use table for function point weights

		Unadjusted Function Points
		UFP = number of inputs * 3 + number of outputs * 4 + number of inquiries * 3 + number of logical files * 7 + number of external
			interface files * 5

		1. Data communications
		2. Distributed data
		3. Perf. Criteria
		4. Heavy HW use
		5. High transaction rates
		6. Online data entry
		7. Online updating
		8. Complex computations
		9. Ease of installation
		10. Ease of operation
		11. Portability
		12. Maintainability
		13. End-user efficiency
		14. Reusability

		Used to come up with Technical Complexity Factor (TCF)

		TCF = 0.65 + 0.01*(sum of 14 tech comp. factors)

		Results in values from 0.65 to 1.35

		FP = UFP * TCF

		Assume average effort is measured as 25 function points per person month

		Effort = x function points / 25 function points/person month

		Can be hard to measure over different organizations, languages, etc.

		There is a working group IFPUG for this topic

	OO Function estimation

		1. Est. the # of classes
		2. Categorize UIs
			No UI		2.0
			Text-based	2.2
			GUI		2.5
			Complex	3.0
		3. Compute #UI classes = # of non-UI classes * UI categorization
		4. #classes = #UI classes + # of non-UI classes
		5. Multiply the total number of classes by a value between 15 and 20 to determine the number of person-days required

	3 tasks , 10, 15, 30 person months
	Calc bcw
	Effort so far 
	Budgeted cost of work (BCW) - estimated effort for each task
	Budgeted cost of work scheduled (BCWS) - sum of estimated efforts sum(BCW) at date
	Budget at completion (BAC) - sum (BCW)
	Budgeted cost of work performed (BCWP) - sum (estimated effort for completed tasks) on particular date
	Actual cost of work performed (ACWP) - sum (actual effort for completed tasks) on particular date

	Earned Value = BCWP/BAC
	Cost Variance = BCWP - ACWP
	Schedule Variance = BCWP - BCWS

	Evidence Based Scheduling
		Base scheduling upon past experiences