Assessing User Interest in Web API Recommendation using Deep Learning Probabilistic Matrix Factorization

Internet 2.0 Things connected to the Internet not only manage data supply through devices but also control the commands that flow through it. The communication technology created by the desired sensor is used by a new computing model so that the collected data appears in Web 2.0 for management. In addition to enhancing Sense efficiency through the simple IoT computing process, it is used in many cases for example video surveillance, and improved and intelligent manufacturing. Every fragment of the system is carefully continued and supervised in this process by software collection using a large number of recurs. An important process for this is to access web APIs from various public platforms in an efficient way. The use of different APIs by developers for the integration of different IoT devices and the deployment process required for this is unnecessary. Obtaining configured target APIs makes it easy to know where and how to get started with the workflow approach. Rapid industrial development can be achieved through this powerful API approach. But finding adequately powerful APIs from a large number of APIs has become a great challenge. However, due to the massive spike in the count of APIs, combining the two APIs has now become a major challenge. In this paper, for the time being, only the relationships between users and the API are considered. In this case, they had to face difficulties in extracting contextual value from their interpretation. So better accuracy could not be obtained due to this. The consequence of the user's time aspect on the cryptographic properties concerning the information collected from the API contextual description can be enhanced by the Deep Learning Probabilistic Matrix Factorization (DL-PMF) method, which improves the accuracy of the API recommendation in considering the cryptographic features of the user in the API recommendation. In this paper, we have used CNN (Convulsive Neural Network) for web elements such as APIs, and LSTM (Long-Term and Short-Term Memory) Network, which works with a diligent mechanism to find hidden features, to find hidden features that suit the tastes of the users. In conclusion, the combination of PMF (Probabilistic Matrix Factorization) evaluation of the recommended results was obtained as described above. The combination of DL-PMF method experimental results was found to be better than previous PMF, ConvMF, and other methods, thus improving the recommended accuracy.