elbraulio / ros_gh Goto Github PK

with small doubles the rank returns NaN, so it should be replaced as 0.

Double.isNaN(rank) ? 0d : rank;

also Aspirant is returning wrong ranking, must be

this.ka*0.75 + this.da*0.25;

the correct Math is

    private double vectorSpace(int i, int j, double[][] rut) {
        double sum = 0d;
        double length = 0d;
        for (int k = 0; k < rut[i].length; k++) {
            sum += rut[i][k] * rut[j][k];
            length += rut[i][k];
        }
        return sum/length;
    }

must be normalized by doc length like this

            length += rut[i][k];
        }
        return sum/length;

we want to replicate Devrec described in this paper. It will be implemented on the examples branch using tools committed on master branch. There is an important difference between Zhang et al. implementation and ours, it is that we are looking for someone to answers a question instead of participating on a project.

All the following quotes were extracted from the original paper.

Data extraction

• we use this data already extracted with ros_gh and available here.

Developer Recommendation Based on Social Coding Activities

• UP Connector: This part is to create the association matrix of users and projects based on the activities in GitHub. Here we get a two-value matrix Ru−p, where 1 stands for participation and 0 stands for the opposite.

• User Connector: This part is to calculate the association between users based on the user project association matrix using Jaccard algorithm.

• Match Engine: In this part, we calculate the association between users and projects according to the user association matrix Ru−u. If we use UAp⟨u1,u2,...,un⟩ to represent users that have already participated in the target project p, we can obtain the match score of each user towards project p using:

Developer Recommendation Based on Knowledge Sharing Activities

• Relation Creator: In this part, we calculate the user tag association matrix. Here we use TF-IDF method. If we use U{u1,u2,...,un} to represent users in StackOverflow, Tu = {t1,t2,...,tn} to represent the tags that related to user u, and C(t,u) to represent the number of times tag t relates to user u. Then we can calculate user tag association matrix using

• User Connector: After obtaining the user tag association matrix Ru−t, we calculate the association of users using Vector Space Similarity algorithm.

• Match Engine: The same as the match engine part in DA-based approach.

Launcher is too general, it might be named as DataExtraction or similar.

it should be

and Javadoc should be

check the first aspirant who math at least the half of tags from question.

Also rename DefaultAccuracy to StricAcuracy

the result of querying how many times a given tag is related to all users has different results if it is queried for each user using FetchTagCount compared with using a single query:

select sum(count) as count
from ros_user_tag
where ros_user_tag.ros_user_id in (select ros_user_id from linked_users) and
      ros_user_tag.ros_tag_id = ?;

There are 1277 tags that looks cut, e.g. :

You can get the complete list with this query:

select *
from ros_tag	
where name like '%...'

fix bugs and devrec implementation.

• create issue release.
• create new branch
• update pom version.
• make a pull request.
• close milestone when pr is acepted.
• make a release on github with pom version.
• check Travis and JitPack build results.

Add this badge to README.md to receive thanks from users.

https://saythanks.io

it is not suitable to have examples within the project because they have not have unit test and increments the project complexity. These tools must be provided by the project but without containing examples inside ignored test. This packages must be excluded from the main project and can be included on a examples branch:

• launcher

• GithubInfoTest

• ignored test from FetchUsersPageListTest

• FetchAnswersTest

• IteratePagedContentTest

• ParticipantsTest

it seems that they are not being used, then they must be removed.

artifact id : rosgh
group id : com.elbraulio

the script for reset db used in test are not windows Multiplatform compatible. Then those test fail when they run on windows.

When indigo or hydro are being processing , it throws an exception for the ar_sys package.
Will bring more details in future ...

Like builds, all we want to know how the extract process made success or failed. It is important to have logs in order to identify errors and possible missing data cases.

actually is sorting asc. Even the test sort asc.

it should be

<dependencies>
    <dependency>
        <groupId>com.elbraulio</groupId>
        <artifactId>ros_gh</artifactId>
        <version>{version}</version>
    </dependency>
</dependencies>

<repositories>
	<repository>
	    <id>jitpack.io</id>
	    <url>https://jitpack.io</url>
	</repository>
</repositories>

README.md must include all versions of data extracted. This info can be added at the bottom of the document with change review.

currently Devrec is out of artifact domain com.elbraulio.rosgh. That is confusing when you import this as

import examples.Devrec

when it should be

import com.elbraulio.rosgh.example.Devrec

update the LICENSE file up to 2019.

I need more data from the ROS Answers' user, specifically:

1. karma

2. joined_at

3. last_seen_at

4. location

5. has_avatar (or the url to the avatar and NULL if it has the default)

6. description

7. real name (if it exist)

8. age

9. badges (and its count)

I'm particularly interested in the first 4, so if it's it require more effort to get the rest of the list I'd already be happy with havine just those four (karma, joined_at, last_seen_at, location).