That's a basic word2vec cluster demo. You can run it with :

• step1: pip install -r requirements.txt

• step2: python cluster.py kmeans --filename data/alert_pcmline_20190617.csv -o output/pcmdline_3.csv -k 3

Now, you would see the result of predict. Have fun.

In my sence, i count the cmdline from alert datastets then cluster it.

select srv_cmd, count(*) srv_cmd_cnt
from xxxxxxxx.xxxx_alert_log_xxxxx
where ds = max_pt("xxxx.xxxxx")
group by srv_cmd
order by srv_cmd_cnt desc
limit 10000;

so that i can find the problem from millions alerts. Actually, it was useful.(also still have some problem)