Project acquisition_analysis

This project encompasses a library for collecting L---e--- profiles, extracting relevant employment data from them for the analysis of job transition behaviors in the context of a Managment Research at Chair for Technology and Innovation Management @TUM. Presentation slides (updated 25/09/2020)

Following are the steps for the complete pipeline of the process:

Module scraping

The module is used to collect L---e--- profiles. It also includes tasks that need to be done before and after the profile collection.

Note: Please configure the corresponding paths and parameters in the file ./config.py before running any submodules.

1. Create an account list for manual sign-up at L---e---

To obtain full information of profiles, we need to be able to log in the platform. Therefore, data used for signing up fake L---e--- accounts should be created beforehand. The submodule scraping/account.py can be used for this purpose.

Configure the ACCOUNT_PARAMS parameters for scraping/account.py in file in ./config.py and run the following to create data for a set of accounts:

python -m scraping.account

2. Use Account Signup Assistant (experimental)

Currently the module scraping/account_signup_assistant.py does not work well since it gets detected by L---e--- during the signup. The result is we get a phone verification instead of CAPTCHA. For this reason, manual signup is preferred by the account data created in the previous step.

Parameters : ACCOUNT_SIGNUP_ASSISTANT_PARAMS

Run:

python -m scraping.account_signup_assistant

3. Filter out duplicate profile URLs

The URLs to profiles should be preprocessed to remove duplications within each URL file (intra-duplicates) as well as between them (inter-duplicates).

Parameters: URL_FILTER_PARAMS

Run:

python -m scraping.duplicate_url_filter

4. Collect profiles

Run the scraper to collect profiles and save them as HTML pages into local storage

Parameters: SCRAPER_PARAMS

Run:

python -m scraping.l---e---_profile_scraper

5. Validate the results

Validate whether there are missing profiles and faulty profiles in the set of profiles that have been scraped.

Parameters: VALIDATOR_PARAMS

Run:

python -m scraping.validator

Module analysis

6. Extract relevant employment data

After collecting the raw data, which are HTML profile pages, we need to extract the employment status from them for further analysis and create the so called intermediate data or extracted data.

Parameters: EXTRACTOR_PARAMS

Run:

python -m analysis.extractor

7. Inspect job transition patterns by acquisitions

By using the extracted employment data from step 6., we now can compare that with the acquisition data and inspect job transition patterns of employees when their companies are acquired. The results will be stored in CSV format for visualization in the next step.

Parameters: INSPECTOR_PARAMS

Run:

python -m analysis.inspector

8. Visualize the results

Graphs to visualize the results of the analysis in the previous step are shown in ipython file analysis/Analysis for Job Transition Pattern by Acquisitions.ipynb. Please use 'jupyter notebook' or 'jupyter lab' to run the file.

Appendix

To debug or visualize the employment timeframes together with the respective acquisition for each employee, please pickle a list of timeframes when running file analysis/inspector.py and then use the ipython notebook analysis/Timeframes Visualization.ipynb to load the pickle files to see the visualization.

Parameters: TIMEFRAMES_VISUALIZATION_PARAMS