Comments (7)
Thank you for opening this issue.
Without any further knowledge, I would be very careful of using the correlation score when the PPS is 0.37 in one direction but 0 in the other. This obviously means that there is only real predictive power in one direction and there is rather no symmetric (two-way) relation between the columns. (if the scores are valid and there was no error)
What you should do is plot the data and have a look at the actual relation.
Can you maybe share your data and the analysis? Then I might have a look at it.
If you need a single score for knowing what the direct two-way relation between the columns is, then PPS might not be the right score for you because it does not give you this direct interpretation and it is not suitable for this two-way, symmetric approach.
Can you maybe state why you need this directly proportional relation? What is the business use case or interpretation?
from ppscore.
Hello Sir,
Thank you for the detailed answer. Actually what I am trying to find is relation between no of words in a post to comments, likes, shares and views. So for that I need to know either they are directly proportional or inversely proportional.
What I am getting from your reply is that if ppscore from x to y is 3.7 and y to x is 0, So this mean they are directly proportional to each other? If I am wrong please correct me.
from ppscore.
Is it possible that you share some data?
Why do you want to find the relation? What do you want to achieve? What is your overarching target? Do you want to predict the likes, shares, comments, views of posts?
from ppscore.
Actually what I want to know if with increasing the no of words in a post weather the no of likes, comments, views etc also increase or decrease? is there any relation between them.
Either they increase directly on inverse with each other.
The data is very sensitive to be shared but they are just numbers. something like views = 2000 , no of words in a post =115. around 20,000 of this type of data.
from ppscore.
If it is possible, please share the (anonymized) data via a CSV e.g. via uploading to a Google Drive - so that I can have a look at it.
If this is not possible due to confidentiality, you can reach out to 8080labs.com for consulting. Otherwise, I am afraid that I cannot help you
from ppscore.
Hello Florian,
You can check the data here.
https://docs.google.com/spreadsheets/d/1bvVXJP__eHmiX7KtiPp211Slor6JGQOBt8bSF3UrPRk/edit?usp=sharing
from ppscore.
Hi, thank you for sharing your data.
I had a quick look at it. When I tried it, the PPS was always 0 which might be caused by the small amount of data. I also plotted it in various ways and also tried to use some binning and it made sense to me that the score was 0 because there hardly exists any pattern
Some observations:
- there are only few posts which have many views - most have an average number of views. Some more, some less but usually in the same range
- Most posts have between 75-200 words. Only some have below 75 words
- There seems to be no clear relation between number of words and amount of views
Asymmetry:
- When a post has a lot of views, it is a little bit more likely to have 140-200 words. However, when a post has between 140-200 words, then there is hardly any difference to posts which have 75-140 words. And actually, posts with 140-200 words have a slightly higher ratio of posts with many views but also a slightly higher ratio of posts with almost no views. So, this case "polarizes" the data into the extremes. However, only very slightly so.
Disclaimer: I would be careful with the findings because those might not be statistically relevant. In case those relationships would be relevant, the PPS could have found them. Since the PPS did not find them, this indicates that the relationships are not strong enough given cross-validation. However, on your bigger datasets, the PPS seemed to be bigger than 0, so the patterns might be valid.
Summing up, there is hardly any relation to find here and it is definitely not what you hoped for (with more words there are more views or inversely). The scenario and data is so complex, that I cannot answer you more questions about it for free. If you want you can try to hire 8080 Labs for consulting but I am afraid that this is out of scope for discussions in the context of ppscore
from ppscore.
Related Issues (20)
- Data preprocessing and information leakage HOT 14
- [SUGGEST] Release a verson supported GPU HOT 2
- ppscore when model_score>baseline_score HOT 3
- There should be an option to override the attribute type like PyCaret HOT 4
- Scikit-learn dependency < 1.0.0 HOT 14
- [Suggestion]: Plot the Decision Tree for pps.score HOT 3
- Readme / docs unclear about using ppscore on time series data HOT 3
- pytests failing with pandas==1.4.0 HOT 1
- Thought on a possible enhancement of the PPS HOT 2
- What does PPS score? HOT 4
- Add support to release Linux aarch64 wheels HOT 4
- Cannot install ppscore HOT 1
- Your package isn't compatible with scikit-learn 1.0.1 HOT 2
- How to report PPS HOT 1
- Question About Data Order HOT 12
- y predicted values given x HOT 3
- Performance HOT 3
- differnt baseline scores for the same y HOT 1
- How to deal with heavy imbalanced data? For example, when the target is 99 "negative" to 1 "positive"
- pandas >2 support HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from ppscore.