We are final year computer engineering student making project on human resource development and management. We want to implement artificial intelligence in our software. we are having very vast emplyoee data. But we dont have any idea about AI. for eg. if we input 10 employee names then the team leader should be given by the software depending upon knowledge base. we are bounded to some time limits so we request you all to help us as soon as possible.

Perhaps a good place to start is the resources page of this website.

I'd recommend against attempting to implement AI when you know nothing about it and have limited time. It will take you months to understand anything appart from basic concepts of AI.

i dont understand where you need AI in your HCM program...
there are many different "AI"s (neural nets, pathfinding, genetic algorithms).
what exactly are you trying to do ?

Your problem description is pretty vague, but I would opt for an expert system in that case.

I think your problem is not about AI. You don't have to use ai algorithms.You should use ordering algorithms.
You should order people with criterion such as success , intelligence ,abilty...If 10 people was given to your sofware , program should select the team leader with the important criteria .If you ask your program for advertisement department , your program can select the leader who has the top creative rating ..
I think it is not easy but in other case ,AI is much more harder solution to your problem.

This problem set is a best fit (judgement) evaluation function.

You have to have some kind of metric (measurements) for different attributes of
the candidates (employees)

You need then to build equations to calculate the weights of importance of each attribute for a particular job type (type would include seniority of position).

You basicly have an equation using the candidates attribute metrics as input and
the output will be a number. You then would run all the candidates in the group
thru the equation and select the highest value as the 'best'.

You may want a minimum threshold to determine if no candidate in the group qualifies for the position.

The equation will be a sum of weight functions each of which has its attribute
as an input. The function may be a curve (usually a histogram - lookup table) hand built to be an interpretation of increased weight


Note- each job has different weight adjustments (functions) for different attributes to reflect the significance for the particular job (leadership may be more important than creativity for a 'teamleader' and technical knowhoew may be important but to a lesser extent).

Notice though that you have to come up with some way to evaluate a candidates attributes to reduce them to a number range (domain for the adjustment weight functions). This will likely be subjective but it has to be done systematicly to give a good measure of the attribute.

The number of attributes you measure is also subjective.

Adjusting the weight outputs of the functions likewise is subjective (for ech position type). The data basicly is a lookup table/array using the input as the index and the array elements being the output.


I would recommend a 1-5 scale to start for the attribute (inputs) and an output
of 0-100 weights from each function (that allows a non-linear function to express extra importance of a particular attribute). Have at least 5 attributes
measured (more if you can define ones independant of each other, but if there are too many the start overlapping).


A variation may be to allow negative weights of certain attributes that badly fit a job position (to lower the candidates evaluation).


The coding is fairly trivial and as is usually the case in AI its programming
the weight data which is most of the (hard) work. It often takes quite a bit of adjusting the evaluation fumction data to balance between factors to get the overall function to give the desired results.

Determining the metrics for the attributes may be the hardest part (figuring out how to measure a persons skills is not an easy task). Manually evaluating all the candidates is also alot of work.

Visual Output can help while tuning the functions (make it easy to visualize the factors and the results) and as a final operational output. writing the data as .CSV could make it easy to read via a spreadsheet application (especially if it can easily turn the data into a graph....)


Do tests against a small sample with known results to see how accurate the evaluation is. All results should be checked to see if there are abberations in your evaluation 'logic'.