im currently doing lstm-ae to detect leak in control valve. i trained the normal data (some from industrial and simulation data), then i test it using industrial data that has anomalies. so the output coming out from my lstm-ae would be: normal abnormal then i feed into a randomforest classifier and train it using leak data. and call back from the output of lstm-ae (normal & abnormal), to classify that these data is consider as leak or no leak. questions: does my quality of data is acceptable since i mixed the industrial and synthetic normal data together to be trained in lstm-ae? i didnt get the correct predicted labels for the classifer when i compare with the ground truth of the actual leak data. do i need to train leak and no leak (normal) in randomforest classifier? 4.please give your opinion on my research study, thank you!! im expecting to get good response and understand better my ltsm-ae model Continue reading...
