made stuff a bit more userfriendly

main.py

@@ -1,14 +1,21 @@
+
+
+#data loading
 import numpy as np
-
-from stroo import *
-
-
 f=np.load("useragents.npz")
 
-model=train_model(f["train"],n=2)
 
+from stroo import train_model
+
+#understands pure strings
+#definitely not the best possible model
+#so if this is at least >0.55 (on more complicated data) we should be able to do something with it
+#also migth be included into an isolation forest
+model=train_model(f["train"],n=3)
+
+#calculate auc. Has disparity between normal, abnormal: np.mean(testy)~=0.065
 print(model.eval(f["testx"],f["testy"]))
-
+#in my tests reaches >0.99
 
 
 

requirements.txt

@@ -0,0 +1,4 @@
+numpy
+keras
+tqdm
+sklearn