In out-of-distribution (OOD) data detection inputs applied to a deep neural system are identifying the untrustworthy predictions. OOD information for the most part alludes to information that’s distinctive from the information utilized to prepare the neural system show in a common sense. The information collection in OOD information a distinctive beneath diverse conditions and for distinctive tests than the real information which was initially utilized to prepare the mode. In this paper, we have prepared an LSTM classification demonstrate to anticipate the activity signals upkeep work sorts done utilizing content portrayals. We have utilized five steps technique beginning from bringing in and pre-process information, isolated ID and ODD information, change, the content information to numeric arrangement, make and prepare LSTM demonstrate utilizing ID information and at long last develop a discriminator and to compare dispersion scores.