Multipath routing uses several paths to distribute traffic from a source to destination. This not only improves network performance but also achieves load balancing and fault tolerance. Though multipath routing is not deployed widely in the internet, current systems chooses the paths with the same lowest cost to the destination and the same administrative distance.

In this paper we propose a flexible multipath network wherein multipath routing algorithm is invoked when the quality of service is affected. Our proposal provides a method to gradually shift from a single path existing network to a reliable multipath network in the future.

We have used machine learning approach to identify the instance when network load balancing is required with good results.