Hi everyone ,
I am a new user of Knime , I had like 5 hours of training with a teacher ( Simple clustering and prediction ) , and I have got a project where I am using Knime to detect anomalys from a data base .. Let me explain briefly the project : it's a wheater station , and I have different datas ( temperature , luminosity .. ) depending on dates , and I couldn't realize a program that will detect the errors .. We tried to simulate some errors and include them in the data base but i can't make a program that will get rid of them .
I have some ideas but i don't know how to execute them .. I was thinking of calculting like the difference between two straights measures , and if the difference is too high , the measure will be detected as an error and erased from the data base ( CSV format ) .. I already used some row filter , to filter like the impossible values as 0% of humidity , however , I have tried using differents nodes such as Numeric distances combined with K medoids , Similarity search and many other stuff .. and I can't get it.
If you have an easier way to detect the anomaly , i would be thankful ! And i am a big beginner in the data analysis field.
Thanks for reading and for your comprehension,
A knowledge seeker !