Just couldn’t sleep last night so decided to do something in KNIME.
I’ve built a workflow (contains Selenium nodes) in which users are scored and ranked based on received likes, replies and like per reply (rate):
I invented a system of scoring: All 3 parameters (likes, replies and the rate) are multiplied by a factor which is calculated by dividing 100 to the max value of the parameter in each period and the score is the sum of these 3 values. This way all the scores in all periods are comparable and all parameters have the same weight (true weight).
Maybe such a ranking will encourage users to be even more active.
Congrats to @Iris, @mlauber71 and @gab1one the top 3 forum users of all time!
Best,
Armin
* I applied a few improvements to the workflow:
Now the workflow takes the top 50 users in the “Replied” field as well. (previously it was taking only the top 50 in the “Likes Received” field)
A “Missing Value” node is added to convert the missing values for rate column (when the number of replies are zero) to zero. So now if the user has a few likes but no replies, he/she still gets the score.
The total scores are now divided by 3. So the scores now scale from 0 to 100 and can be compared easier. When I say the scores are comparable: For example when @mlauber71 has the score of 69 in the month field and @armingrudd has the score of 61 in the quarter field, it means @mlauber71 has done better in this month than @armingrudd in this quarter.
Hi all!
Nice one @armingrudd. I just hope you feel asleep after this easily
Don’t quite get the scoring system but according to names seems it is quite accurate @gab1one that means your final thesis is done?
Have a nice day,
Ivan
Actually I’m very happy that such an idea came to my mind. Such a mid-night idea. If you investigate the scoring system you will find it very interesting. There is a factor which is calculated by dividing 100 to max value for each parameter (e.g. likes). Each parameter is then multiplied by this factor and the score would be sum of the values.
In shorter periods where likes and replies are lower but the like per reply (rate) may be high, this factor is high for likes and replies and low for the rate and as the period gets longer this behavior changes to become vice versa. So for all the periods we have scores that can be compared to each other and all the 3 parameters have a true equal weight.
For example, @Iris has about 2k replies in all time period and about 200 likes (which makes the rate about 0.1). As in all time period the number of replies are very high, the factor for this parameter is low (0.05)
and as the rates are low the factor for this one is high (62.5). So, more like per reply (higher rate) in all time period will give you much more score than the other parameters. The factor is actually balancing the values.
No, as I was excited building something I didn’t sleep
Really great work 
. Unfortunately I couldn’t make it to the top 10 (I’m 12th). 