Predicting human behavior from social media using mRMR with COA

Social media users can participate increasingly by sharing online information, and the work in this research content can be helpful to assess their personalities. Personality prediction is defined by extracting the digital features from the digital content and mapping those features into a personality prediction model. This human behaviour identification will be helpful to multiple job processes. The advancement of data-based approaches in social sciences will be helpful to model human behaviour based on unstructured text data. Due to the simple nature of the big five personality traits, it has been used in analyzing human behaviour. This paper focuses on predicting human behaviour based on personality prediction with unstructured textual data mining. So far, many researchers have proposed a personality prediction model based on deep learning approaches. However, the existing model slack processing time and the ability to capture the real meaning of the word. This paper proposes a deep learning-based prediction model from the data stored on Social media such as Facebook, Twitter, and Instagram to overcome these issues. Initially, the data are preprocessed to remove the irrelevant data such as URL, symbols and stop words. The features are extracted using the proposed mRMR based cat optimization algorithm from the preprocessed data. This approach identifies the relationship among feature sets and traits from datasets. The human behaviours are classified with an Improved LSTM classifier optimized with a forest optimization algorithm. The proposed mRMR-Cat optimization-based feature extraction and LSTM with forest optimization approaches outperform all feature extraction average baseline sets and Classify on multiple social datasets with improved accuracy of 86.5%, 88.4% and 90.16% for the datasets Facebook, Twitter and Instagram.