Современные информационные технологии/ 3.Программное обеспечение

Stradetskyy O., Bilas O.

Lviv Polytechnic National University,Ukraine

Software project estimation time error calculating

In programming as in everyday life, often need to estimate how long it takes to perform a particular task. It is necessary for the success of the project and is one of the hallmarks of transparency and understandability of the system. After all, if one cannot describe what you need to do to achieve the goals of the project, then it may not properly assess the time it is expected to spend on the task. This estimated time is approximate and in most cases are not precise. Although, this assessment is often required to make during the development of software as a programmer that performs tasks on the project and the project manager is responsible for all software. The customer wants to know the least about how long it takes for developer to complete the task.

Since managers are often not very transparent imagine the amount of work before they start, as well as the rapidly changing requirements of software that practiced methodologies Agile - there are many risks that cannot be taken into account when assessing planning time task or project. It is advisable to clearly identify the stage of planning and risk assessment.

In order to assess as accurately as possible the planned time to be fully divide the task into smaller parts so long until you clear understand idea about each of the items. Often in practice the approach when the task is divided into paragraphs so long until you evaluate it in bounds. But in this case, the person who counts the scheduled time may not consider some risks as random and periodic. So should propose and investigate algorithms for error evaluation scheduled time, and develop software to help calculate this error for certain tasks.

To estimate the error of the planned time is divided into two basic ways:

· Estimation total error

· Estimation of error based on earlier experience.

The estimated total error (standard deviation, designated as or ) - in probability theory and statistics, the most common indicator of dispersion of the values of a random variable with respect to its expected value. In fact, if we applied problems, the standard deviation - is the most used indicator of variability of the object, which shows how much the average individual values deviate signs of their average ().

That is in our case, this assessment will show how much the average current estimate deviates from the planned time average.

Standard deviation - is equal to the square root of the variance of the random variable [1]:

According to the formula of calculating the variance:

· With small sample (n<=40-50) Bessel correction is [2]:

·

Where: - standard deviation, constant average score standard deviation of a random variable relative to its expected value;

— variance;
— i-th element of the sample;
— the arithmetic mean of the sample:

— the volume (size) of the sample.

So, in order to calculate the standard error, use the following formula [3]:

where, – standard error

– standard deviation, unchanging rating average deviation of a random variable with respect to its expected value;

– the number of elements in the sample;

– sample standard deviation

— i-th element of the sample;

— the arithmetic mean of the sample;

— the volume (size) of the sample.

Consider the example calculations, even made 5 guest scheduled time (5, 20, 40, 80, 100)

Looking standard error,

So we can assume that the error is 18 hours, which means that it is worth to total 245 hours to add 18 and then the total number of hours would be 263.

However, this method ignores a very important factor that can significantly affect the performance results and the experience of the person who will perform these tasks. It is therefore advisable to develop an algorithm that would also give error estimates for the planned time, but based on earlier experience.

Method error estimates based on earlier experience, based on a simple and reliable principle. Based on indicators obtained during the preliminary work performed, calculates the coefficient of error,

where: — і-th element of routine sampling time; — і- th element of the sample spent time;

Next, to estimate the error () for the planned time a new task (), you must multiply by a factor of ().

where: - the number of hours, how many rejected theoretically predictable at times, that costs of its implementation.

Example of calculation:

Planned time,	5	20	40	80	100
Spent time,	8	16	52	78	125

This means that the assessment of the planned time of 48 hours should be increased to 67 hours.

After analyzing these two algorithms, we can say that they complement each other, although to some extent and overlap each other. It should be noted that in practice, both methods should be used to analyze performance of one and another, and even combine them. This can give good results and reduce the risk of exceeding the scheduled time.

References

1. The Cambridge Dictionary of Statistics, CUP / Everitt, B.S. – 2003

2. Kenney J., Keeping E.S. Mathematics of Statistics / van Nostrand.- 1963.- p. 187

3. Zwillinger D. Standard Mathematical Tables and Formulae.- 1995