The site description of Cross Validated says:

Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization

And the site description of the Data Science SE site says:

Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field.

But, as ML is already a topic in CV, there are significant overlaps in the questions.


  1. Using monthly product usage data to predict customer churn

  2. Modeling customer churn - Machine learning versus hazard/survival models

And there are a plenty of such similar questions. So, is a separate site (Data Science SE) required, when the majority of the questions are on-topic on the CV site?

Edit: This question is not about how the sites are faring, it is about the key differences, which sets them apart and identify them as two separate sites so that the tour/about pages can be edited accordingly for avoiding future confusion.

I have found this discussion, but the reasons and explanation given, is still wanting as to prove that a separate site is needed for data science apart from CV and SO.

  • 1
    Data Science need to add some info to their "on-topic" section: datascience.stackexchange.com/help/on-topic – James Sep 9 '15 at 14:22
  • I have quoted the above, from the tour pages of both. (datascience.stackexchange.com/tour). And if the questions are observed, bulk of the questions in DS SE are on-topic in CV. – Dawny33 Sep 9 '15 at 14:28
  • @James Is it okay if I reproduce the same question in the CV and DS meta sites? Just in case, I might get more answers from the users there. – Dawny33 Sep 9 '15 at 14:38
  • Is cross-site posting allowed. You'll get enough attention here, just give it some time (at least 2-3 hours, but more really as people get home from work + different time zones) – James Sep 9 '15 at 14:49
  • 3
    There's no sense in talking about DS.SE. According to the stats it is almost dead. Stats.SE (CV) is another matter - there are upwards of 30 active experts each week (data.stackexchange.com/stats/query/342741/…) – Deer Hunter Sep 9 '15 at 18:09
  • Why do you want someone to prove the need for DS? It shifts the main thrust of your question considerably, and for goodness sake, you can leave the poor DS site to die in peace. – Deer Hunter Sep 11 '15 at 9:38
  • 1
    @DeerHunter It is not about proving. I want to know the main differences which would set them apart. And I don't think the discussion link in my recent edit properly addresses it. – Dawny33 Sep 11 '15 at 9:41
  • 3
    Now we have the AI SE site as well, and I am lost where to go! – A.Rashad Aug 25 '17 at 13:51

This is rather a long comment... (This was originally written for CV and Data Science: semi-identical twins?, but then I found this question).

Comparison by tags

One way to look at datascience.SE (DS) / CV is to compare the sites by tags.

Top tags of DS on 2019-08-11:

  1. machine-learning: 5881 (vs machine-learning: 12819 on CV)
  2. python: 3219 (vs python: 2529 on CV)
  3. neural-network: 2511 (vs neural-networks: 4711 on CV)
  4. deep-learning: 2338 (vs deep-learning: 2077 on CV)
  5. classification: 1623 (vs classification: 4859 on CV)
  6. keras: 1385 (doesn't exist?)
  7. scikit-learn: 1084 (vs scikit-learn: 1133 on CV)
  8. r: 1011 (vs r: 20394 on CV)
  9. tensorflow: 997 (doesn't exist?)
  10. nlp: 954 (doesn't exist?)

Top tags of CV on 2019-08-11:

  1. r: 20394 (vs r: 1011 on DS)
  2. regression: 18727 (vs regression: 726 on DS)
  3. machine-learning: 12819 (vs machine-learning: 5881 on DS)
  4. time-series: 9621 (vs time-series: 835 on DS)
  5. probability: 7606 (vs probability: 159 on DS)
  6. hypothesis-testing: 6468 (doesn't exist?)
  7. distributions: 6212 (doesn't exist?)
  8. self-study: 6171 (doesn't exist?)
  9. logistic: 5170 (doesn't exist?)
  10. bayesian: 5048 (doesn't exist?)

One can see a couple of things here, I think:

  • CV is much bigger than DS, by now. Considering the fact that CV is 6 years older than DS (source and source), I guess this is natural. It would be interesting to get some data for stack exchange sites to try to predict the growth :-)
  • DS seems to attract more people from computer science, where CV seems to attract more people from mathematics.

Further analysis

It would be interesting to get a graph of the tag growth by month for both sites (in one graphic).

I would also like to see which kind of questions get closed on both sites. Which tags do they have? How often does it happen that a question gets migrated DS -> CV and how often CV -> DS?

Discriminating stats.SE from datascience.SE

For me, I can say that I like DS more. The name of the site seems to be more clear to me. Only from the name, I know that this includes machine learning / analysing data / classification / prediction. But cross validated? I know what cross validation is, so is CV only about testing? And why is it called "crossvalidated" but has stats.SE as an URL? This seems unfortunate.

When one likes to have one big site, then why not merge both in math.SE / stackoverflow / cs.SE / opendata.SE...?

I think both sites have a reason to be there. It seems to me that stats.SE should be about statistics. Yes, it has a large number of machine learning questions, but my guess is that this is simply because of the age. StackOverflow also has a lot of ... well, everything. Because it was there first. That doesn't mean the other sites are useless or that new questions shouldn't be moved. One example is the latex tag on SO. Almost all questions tagged with "latex" on SO get moved to tex.SE. Similarly, I think almost all questions tagged with "machine-learning" should get moved to datascience whereas "statistics" is a candidate I would rather see on stats.SE.


import requests

def get_api_result(uri):
    resp = requests.get(uri)
    return resp.json()

def get_toplist(ds, other, ds_tags, cv_tag_dict):
    ds2cv_tagname = {"neural-network": "neural-networks"}
    print(f"Top tags of {ds}")
    for i, tag in enumerate(ds_tags["items"], start=1):
        new_tagname = ds2cv_tagname.get(tag['name'], tag['name'])
        tag_cv = cv_tag_dict.get(new_tagname, None)
        if tag_cv is None:
            vs_string = "doesn't exist?"
            vs_string = f"vs {new_tagname}: {tag_cv} on {other}"
        print(f"{i}. {tag['name']}: {tag['count']} ({vs_string})")
        if i == 10:

base = "https://api.stackexchange.com/2.2"
fstring = "{base}/tags?page=1&pagesize=100&order=desc&sort=popular&site={site}"
ds_tags = get_api_result(fstring.format(base=base, site="datascience"))
cv_tags = get_api_result(fstring.format(base=base, site="stats"))

ds_tag_dict = {tag['name']: tag['count'] for tag in ds_tags['items']}
cv_tag_dict = {tag['name']: tag['count'] for tag in cv_tags['items']}
get_toplist(ds="DS", other="CV", ds_tags=ds_tags, cv_tag_dict=cv_tag_dict)
get_toplist(ds="CV", other="DS", ds_tags=cv_tags, cv_tag_dict=ds_tag_dict)
  • 3
    Neat analytics :) Yeah, you're right with: DS seems to attract more people from computer science, where CV seems to attract more people from mathematics. (+1) – Dawny33 Jan 27 '16 at 8:55
  • I would really love a data-based discussion about this issue. I could imagine an animated graph where one sees how new stack-exchange sites pop up. Tags, when the same, attract two sites. When being different, they are repelled. I guess at first most sites would be very close to SO and then move away. When one knows how this process works, then we can probably speed the process up. (Explorative data analysis is also something I can see on DS, but not so much on CV) – Martin Thoma Jan 27 '16 at 8:59
  • 1
    See ML questions: here or at Data Science?. A summary of my answer is that we can't & shouldn't try to separate Machine Learning from Statistics, & that ML questions have always been positively welcomed on Cross Validated, not merely put up for lack of a better place for them. – Scortchi - Reinstate Monica Jan 28 '16 at 12:26
  • 2
    Please read "[...] merely put up with [...]" in the above. By the way, the idea behind the site name is that we "cross-validate" each other's answers. Never gets stale, just like all those hairdressers called "A Cut Above", "The Final Cut", &c. – Scortchi - Reinstate Monica Jan 28 '16 at 12:59
  • Yea the cross-validated is actually a neat name. – WestCoastProjects Feb 8 '19 at 20:06
  • 1
    I would second about the DS folks being CS-ey and CV being maths-y. i'm on the former side and am still intimidated by the CV site. Trying to work my way out of that situation .. but in the meantime ds site is v comfortable. – WestCoastProjects Feb 8 '19 at 20:07
  • Do you have a few typos in the section on top tags of CV? The parenthetical CV should be DS, I think. – mdewey Aug 11 '19 at 8:25
  • Thank you, @mdewey - I fixed it. – Martin Thoma Aug 11 '19 at 11:36

After getting involved in both the sites; I think I now have enough experience of both, to make a statement.

The DataScience SE is about the problems and doubts which data scientists encounter on a regular basis like "how do you run neural networks on a cluster efficiently?" and "How do I efficiently setup a MachineLearning process on a server", etc do not belong to the CrossValidated or the StackOverflow sites. They can be only answered in the DataScience site.

As of now, a major chunk of DataScience SE's questions are on-topic with posts on CrossValidated; and it is just because due to the fact that statistics make a major role in data science, so it is and will be a common happening across both the communities.

However, the DataScience SE and the CrossValidated sites are completely different from each other and serve their purposes very well.

A similar discussion on the CrossValidated meta.

  • 4
    So can we agree that CV is oriented towards the more academic aspects of data science and statistics (e.g. understanding Singular Value Decomposition, Regularization, etc.) whereas DS is geared more toward implementation and tools (e.g Matrix Factorization Libraries, Google Dataflow, Tensorflow, etc)? – Black Milk Jul 21 '17 at 19:25
  • 4
    The difference between Cross Validated and the Data Science SE seems parallel to the difference between the Computer Science SE and the Software Engineering SE, with a bit of Stack Overflow thrown into both. – Travis Jul 22 '17 at 22:56
  • Thx for this clarification: while it does corroborate my suspicions it is useful to have brought that out to the open. – WestCoastProjects Feb 8 '19 at 20:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .