Anomaly Detection in SAS Viya

[Rahul_B]

Hello,

Need help with detecting anomalies in my dataset 

data hdfs_kim.node_monitoring;

input  date node1 node2 node3 node4 node5 

datalines ;

01-02-2020 0.45 0.44 0.78 0.32 0.99

02-02-2020 0.34 0.32 0.89 0.56 0.77

03-02-2020 0.89 0.65 0.76 043 0.81

04-02-2020 0.73 1.34 0.66 0.33 0.49

05-02-2020 0.23 0.44  0.54 0.66 0.66

06-02-2020 0.88 0.76 2.56 0.61 0.71

;

run;

Can help me find anomalies in this dataset - I have used 

proc rpca data=hdfs_kim.node_monitoring
lambdaweight = 3.5
outsparse=hdfs_kim.sparsemat2;
id date;
run;

proc print data=hdfs_kim.sparsemat2;
run;

proc rpca data=hdfs_kim.node_monitoring
scale center;
id date;
anomalydetection;
savestate rstore=hdfs_kim.store;
run;

proc astore;
setoption rpca_projection_type 2;
score rstore=hdfs_kim.store data=hdfs_kim.node_monitoring out=hdfs_kim.scored;
run;

proc print data=hdfs_kim.scored;
run;

I was not able to interpret results using RPCA  when i have node1 -16, any help with other process or procedure i am willing try.

Thanks 

Rahul 

[Zohreh]

Hi Rahul, 

In your output scoring file "scored", you can see the last column which is labeled as "outlier detection score".

Value 1 in that column indicates that  the scoring observation is outlier.

You can read more about the anomaly detection functionality of proc RPCA here: SAS Help Center: The RPCA Procedure 

Thanks,

-Zohreh 

[Rahul_B]

Hi,

I ve read the whole procedure thoroughly but the scored dataset does gives anomalies but when it comes to more than variable ( multi dimension ) this  doesnt correctly determines where the anomalies are ? 

Any suggestions ?

-Thanks 

Rahul