Features

The University of California, San Diego and FICO (formerly called Fair Isaac, Inc.) are pleased to announce the Sixth Annual UCSD Data Mining Contest. The goal of the contest is to detect anomalies in e-commerce data, as accurately as possible.

TASKS

The contest consists of two distinct but similar tasks. In each task, contestants are given e-commerce data records. Each record is labeled as being an "anomaly" or not. The goal is to train a classifier that can predict whether or not test records are anomalies.

The "easier" first task is to maximize accuracy of binary classification on a test set, given a fully labeled training set. The performance metric is the lift at 20% review rate.

The "harder" second task is similar to the first task, except values for a couple of additional fields are provided in each record. These fields contain information with potential predictive value, but are more difficult to exploit.

RULES

The contest started on May 15, 2009. It ends on July 15, 2009, which is the last day on which predictions may be submitted.

Contestants can enter the contest alone, or as a team of up to four people. Each person may be a member of at most one team. Contestants may be located anywhere in the world, and there are no citizenship restrictions. All current full-time undergraduate and graduate students, and post-doctoral researchers studying full-time, are eligible to compete. Due to the academic nature of the competition, anyone working for a company may not compete.

Teams fall into two categories: teams composed of undergraduates only compete in the undergraduate division, while all other teams compete in the graduate division. Each team may compete in one or both of the tasks.

Once a day, teams may see their current performance by submitting their predictions for the test set. Scores are displayed on the contest leader board.

PRIZES

In total $8000 in prizes will be awarded. For each task and each division $2000 will be awarded:

• First Place $1000
• Second Place $600
• Third Place $400