The problem LFT Pattern Analysis Using Clustering solves

Everyday, thousands of liver function tests are taken in the millions of labs in this country. Analyzing these records and searching for patterns is a very tedious task. This solution makes it easier for lab technicians to recognize patterns in data. These patterns help in laying out a path towards further tests which must be done to find out the cause of abnormality. Depending on the cause of the abnormality further tests can be recommended. The 3 major causes of abnormality are cholestatis, hepatocellular damage and synthetic function. Thus, it greatly reduces the workload on lab technicians who have to sift through hunderds of liver function tests everyday.

Challenges we ran into

As we were asked to follow a clustering based approach, one of the main challenges was to find which clustering algorithm to use. We initially started out with rule based clustering. But, this classified a majority of the records as abnormal which is contradictory to what was actually advised to us. So we switched to agglomerative heirarchical clustering. But due to the sheer size of the dataset we were given, it failed. A similar problem occurred during the DBSCAN clustering. Thus, after trial-and-error we came to k-means algo which gave us the appropriate results and patterns.