The SVM trained on the non-normalized data gave better results than expected. There is a very low probability that this is by chance since the training and test data were chosen such that the a possibility of good results by mere chance is eliminated. Nevertheless, to be sure, it could be tested for a single sensor.
The availability of data is an issue here. The CSN, being a relatively new project, has recorded only a limited number of earthquakes. Also, once a Phidget is unplugged and setup again from a different machine, it is assigned a new identification number. So, it becomes very difficult to figure out a pair of identification numbers which actually belong to the same Phidget. Maybe, in the future, the CSN develops a work around.
The availability of data is an issue here. The CSN, being a relatively new project, has recorded only a limited number of earthquakes. Also, once a Phidget is unplugged and setup again from a different machine, it is assigned a new identification number. So, it becomes very difficult to figure out a pair of identification numbers which actually belong to the same Phidget. Maybe, in the future, the CSN develops a work around.
No comments:
Post a Comment