Name of Project: Job Prediction Using Frequent Pattern Mining
Group Members:
Anukriti Ranjan
Rasna Goenka
Sri Nitya Anupindi
Sukanya Sasmal
Suruchi
Vajha Myna
For every UG course in IIT Kharagpur, we are required to opt for an IT breadth. We chose Softcomputing tools in Engineering. For this course we are required to do a mandatory project submission at the end of the term.
Ours was a pretty heterogeneous group in terms of our major subjects: Electrical, ECE, Biotech and Chemical, we had it all. When the time of deciding the topic of project came, reaching a consensus was difficult. One fine day I suggested we do something on Data Mining, gave them a brief background of the field and everybody agreed. Now what exactly should we do. I had some experience in this field as I had done an internship in IIM-A the previous summer in this field. We started searching about various topics in this vast field and thinking what has already been done and what can be done.
I recalled that one of my fellow-interns was doing work on something called Frequent Pattern Mining. I googled the thing and it sounded interesting. Then I asked him to forward all the papers that he had read for it. I read them all, got even more interested. Then suddenly an idea popped in my mind, we can use pattern mining for prediction based on attributes. Since generally techniques like Time-Series Analysis are used for prediction, the idea was fresh and we did not find any previous work done in this regard. Hard Part Over or so we thought.
Coding part started:
My guide in IIM once said that Data Mining Applications should preferably be Java based. Since Java is a freeware and platform independent. So we decided to code in Java. The fact that 5/6 people in our group did not know Java wasn't helping our cause. We were using Apriori Algorithm for pattern generation. It took us 5 days to get that Algo working. Since there was no previous work done, we had to work out each and every logic and then code it. So even for seemingly small things it took us hours to think and implement.
Bits and bytes of the code was working but till the night before the presentation our dataset wasn't ready. Add to that our GUI was only partially functional, we were in big trouble. We decided to keep awake and code through the night. Sometime around 3 AM the bits and bytes were all assembled into a main code and it was working. Now the only thing left was linking it with the output and running the Apriori code to generate patterns beforehand so that during the demo we only need to run the matching part of the code. We decide to take a break. Anukriti and Suruchi go to JCB. Me and Myna decide to take a short nap.
Time 4:45 am : I wake up with a start. Myna is sleeping. No sign of Anukriti. I think that they got the whole thing working and there is no need to keep awake and I go to sleep.
Time 6:45 am : Knock on my door. A frantic Sukanya informs me our code isnt working. The output couldnot be linked with the input. There are no generated patterns.
Time 8 am: We reach Vikramshila. Start running the code for pattern generation. First file takes 40 minutes, 19 more to go. Time left: 5 hours.
Time 12:30: Lunch Break. 10 files are done. We reduce the minimum support value for the rest of the files and generate patterns. GUI still not working. Giving arbitrary errors none of us is able to decipher in that high tension situation. I and Anukriti frantically go through the presentation
Time 1:30: Presentations resume. We approach sir to give us some time as we didnot have the patterns ready by then. Sir declines.
Our turn came. I and Anukriti start the presentation. The few people who were listening weren't getting anything of what we were trying to explain. Ours was the first group for which Sir asked questions in the middle of the presentation. It was a disaster.
Presentation ended. We were sitting near Vikramshila canteen, venting out our frustration. I was feeling especially guilty for guiding the group through this path. After almost an hour we go back to the presentation room. Brave through rest of the presentations, do the documentation. At the end of the day we decided we should talk to sir and explain again what we did.
To our surprise, Sir himself called us. He told us "Your project has tremendous potential. You write a paper on it. I will send it to an International Conference On Data Mining, if selected atleast one of you will get a chance to go to Germany".
We walk out of Vikramshila, in silence yet smiling ear-to-ear.
True that the presentation was a disaster, but just the fact that Sir understood we had done something new, which is neither Fuzzy nor farji, filled us with satisfaction. And the crappiest of days turned into the happiest one.