Journal of Intelligent Computing Vikas Chaudhary, Mesfin Jariso Delbu, R. S. Bhatia 9 1 2018 http://www.dline.info/jic/fulltext/v9n1/jicv9n1_3.pdf The Self Organizing Map is widely used in classification, vector quantization etc. In SOM, a winner is identified for each input data and weights of winners and its neighborhood are updated. Using the conventional SOM approach, the learning process is influenced by noisy data and as result degradation in learning efficiency. A new approach called Noisy Data SOM (NDSOM) is proposed to identify clusters efficiently using some additional states in the learning process. These additional states control the weight updating process of SOM according to available noise in the input data. The proposed SOM and conventional SOM behavior experiments on various input dataset having noise also. After the study of simulation results, we can conclude that the proposed SOM successfully extracts the cluster and gives better results. Also the proposed SOM preserves the input topology in a better manner with more neuron utilizations.