Most Valued Data Science Skills

---

1. Introduction

In today’s data-driven world, data science has become one of the most in-demand and interdisciplinary career paths. Professionals are expected to combine technical, analytical, and communication skills to extract insights and drive decisions using data.

This project explores the question: “Which are the most valued data science skills?”

Using a dataset from Kaggle (Data Science Job Postings & Skills, 2024), the goal is to identify the most frequently requested skills among employers in data science job postings.

---

2. Tools and Workflow

This project was completed individually using:

• PostgreSQL (pgAdmin 4) — to design and store normalized tables
• RStudio — to connect to the database, perform analysis, and visualize results
• Kaggle Dataset — for raw data collection

This workflow demonstrates data acquisition, transformation, storage, and analysis from start to finish.

---

3. Data Collection and Loading

Dataset: Data Science Job Postings & Skills (2024)

Author: asaniczka

Platform: Kaggle

Source: LinkedIn job postings

License: ODC Attribution License

The dataset includes two columns:

• job_link: a unique link for each job posting
• job_skills: a comma-separated list of skills mentioned in the listing

Import and Load into PostgreSQL

A PostgreSQL database named data_science_jobs was created, and the dataset was imported into a raw table:

CREATE TABLE raw_job_skills (
  job_link TEXT,
  job_skills TEXT
);

---

4. Database Normalization

To comply with relational design principles, the data was normalized into three tables:

• jobs (job_id, job_link)
• skills (skill_id, skill_name)
• job_skills (job_id, skill_id) — linking table for many-to-many relationships

The normalization process ensured a many-to-many relationship between job postings and skills, allowing efficient querying and analysis.

The following SQL script summarizes the normalization process:

-- Create normalized tables
CREATE TABLE jobs (
  job_id SERIAL PRIMARY KEY,
  job_link TEXT UNIQUE
);

CREATE TABLE skills (
  skill_id SERIAL PRIMARY KEY,
  skill_name TEXT UNIQUE
);

CREATE TABLE job_skills (
  job_id INTEGER REFERENCES jobs(job_id),
  skill_id INTEGER REFERENCES skills(skill_id),
  PRIMARY KEY (job_id, skill_id)
);

-- Insert distinct job links
INSERT INTO jobs (job_link)
SELECT DISTINCT job_link
FROM raw_job_skills
WHERE job_link IS NOT NULL;

-- Insert distinct skills
INSERT INTO skills (skill_name)
SELECT DISTINCT trim(skill)
FROM (
  SELECT unnest(string_to_array(job_skills, ',')) AS skill
  FROM raw_job_skills
) AS all_skills
WHERE trim(skill) <> '';

-- Link jobs to skills safely
INSERT INTO job_skills (job_id, skill_id)
SELECT j.job_id, s.skill_id
FROM raw_job_skills r
JOIN jobs j ON r.job_link = j.job_link
JOIN LATERAL unnest(string_to_array(r.job_skills, ',')) AS skill_array(skill_text) ON TRUE
JOIN skills s ON trim(skill_array.skill_text) = s.skill_name
ON CONFLICT DO NOTHING;

---

5. Connecting PostgreSQL to R

The normalized tables were connected to RStudio through the RPostgres and DBI libraries. A secure password management approach was implemented using the .Renviron file, which stores the PostgreSQL password as an environment variable (POSTGRES_PASS). This ensures reproducibility while keeping credentials confidential.

con <- dbConnect(
  RPostgres::Postgres(),
  dbname = "data_science_jobs",
  host = "localhost",
  port = 5432,
  user = "postgres",
  password = Sys.getenv("POSTGRES_PASS")
)

---

6. Data Exploration and Visualization

The query below retrieves the Top 10 Most Frequently Mentioned Skills:

top_skills <- dbGetQuery(con, "
  SELECT s.skill_name, COUNT(js.job_id) AS frequency
  FROM job_skills js
  JOIN skills s ON js.skill_id = s.skill_id
  GROUP BY s.skill_name
  ORDER BY frequency DESC
  LIMIT 10;
")

top_skills

##            skill_name frequency
## 1              Python      4746
## 2                 SQL      4516
## 3       Communication      2481
## 4       Data Analysis      2146
## 5    Machine Learning      1934
## 6                 AWS      1714
## 7             Tableau      1651
## 8  Data Visualization      1528
## 9                   R      1527
## 10               Java      1399

Visualization

top_skills %>%
  ggplot(aes(x = reorder(skill_name, frequency), y = frequency)) +
  geom_col(fill = 'orange') +
  coord_flip() +
  labs(
    title = 'Top 10 Most Valued Data Science Skills',
    x = 'Skill',
    y = 'Frequency'
  ) +
  theme_minimal()

After completing the visualization, the database connection was closed:

dbDisconnect(con)

---

7. Findings

The analysis revealed that:

• Python and SQL are the most frequently mentioned skills in data science job postings.
• Other important skills include Communication, Data Analysis, and Machine Learning.
• Visualization tools such as Tableau and programming languages like R also appear frequently.

These results emphasize that employers value professionals who combine technical expertise with strong analytical and communication abilities.

---

8. Conclusion

This project demonstrates the complete process of:

1. Collecting real-world job posting data,
2. Loading and normalizing it in a relational PostgreSQL database,
3. Connecting RStudio to perform exploratory data analysis, and
4. Visualizing the findings effectively.

Through this end-to-end workflow, I gained practical experience in data acquisition, database design, and data visualization — all key components of modern data management.