I am working on deep learning-based Recommender systems. I want to know if I can use Word2vec library instead of RNN to extract features/information from user session logs and application log files to get help in predicting more accurate item of interest to an end user...
If someone can suggest me a better solution, it will be very helpful .
Word2vec can be used to help turn word-like tokens into numerical features, yes.
Whether it'll help on any particular log data depends largely on the quality/type of info in the logs, and how well that info is related to project-specific goals. For example, are the logs already, or can they be coerced into, something that's at least a little like natural-language texts?
