ചൊവ്വാഴ്ച, നവംബർ 19

Challenge: Can you find out a file is a CDR,TOWER-CDR, CELL ID, SDR and IPDR from the data?

Ok, that's a good question,
First, let's find out whether we can find out the header row of each kind.

In the car of CDR,

Header row must contain date/time or date and time together, duration.

Can you say that the idea will be 100% foolproof?

No, IPDR may contain those fields too, but it's rare that an IPDR file comes with duration field along with date and time in a separate column, and more than that IPDR will contain IP’s for sure.

So no need to explain it further..
If you were able to find text” IP” along with date and time in the header field, it's an IPDR.

What are the left-outs?

Cell ID?
(don't care too much about that buddy, anyway, we will incorporate open-cell id soon )

Still, you can check for latitude, longitude, azimuth and address in header fields to identify it.

You can check for Name, Address, pin code, city, state headers. Can't u?

So we covered all the documents!
It's party time. Isn't it?

Someone may say ”I can see your overconfidence, but it's not over buddy”

”Yeah, I know, I know”

Next, you will ask me, ”How to find out a CDR file is Normal CDR, Tower CDR or IMEI CDR”

That's tricky, I won't say it's easy

Let see, we can figure it out or not

How to know its a normal cdr?
(*conditions apply)
If you can find out A party and B party fields, you can find out by checking unique fields count.
Let me explain it,

Let's create four variables


Let's see the conditions
1. If Unique Tower ID count is approximately one, and it has more than one unique AParty BParty and IMEI count, then it's tower cdr
2. If Unique IMEI count is approximately one and it has more than one tower id or AParty or BParty count then it's IMEI CDR
3. If anyone of the AParty or BParty unique count is approximately one and the remaining entries count are more than one, It's a normal CDR.

I think this approach is more reliable than, finding document type from the file name.