Sentiment analysis is the process to analyze the opinions, emotions or sentiments regarding a review, comment, organization, firms or some event etc. with the introduction of social media people tend to share their sentiments, opinions, emotions and ideas through them. Urdu which is a dominant language in Indian sub-content. Most of people tend to share their sentiments using Urdu as one of main language on these social media sites. In this paper the Urdu text available on different social media platformswill be distributed into their vector forms by using the Word2vec model and Long Short-Term Memory Units will be utilized for text classification and SoftMaxfunction will be used as an activation function in LSTM. This SoftMax function has been used for creating sentence polarity of positive, negative or neutral attribute. In this resear ch work the whole process has been used with recurrent Neural network.