Any kind of business depends on customer feedback. Best resource of such feedback is online platforms where user share its experience. Analysis of such reviews play an important role for the product, service improvement. This paper has developed a sentiment class identification as per user text reviews. In order to improve classification feature set were optimized by use of weed optimization algorithm. As per dynamic nature of algorithm no prior training or knowledge need to be used for feature optimization. Selected feature set were transform into numeric form. Transformed feature set were used for the training of linear regression model. Experiment was done on real dataset of user text reviews. Result shows that proposed weed and deep learning model has increased the sentiment detection model accuracy.
Keywords:-Sentiment mining, Digital review analysis, Text mining, Feature Extraction.