15 Data Science FYP Ideas Using Pakistani Data (2026)

A data science FYP is judged on the question and the data as much as on the model. Pakistani data gives you both: a question people here care about, and cleaning work that shows real skill. These 15 data science FYP ideas each say where the data lives, how much cleaning it needs, which method fits and what you hand in at the end.

Expect PDFs and messy Excel files. Turning them into a clean dataset is part of the work and deserves its own chapter in your report. None of these repeat the data science ideas in our main list of 70 FYP ideas for computer science students.

FYP ideas by degree: AI · software engineering · IT · cyber security · web development · mobile apps · Python · computer engineering (IoT) · business (BBA) · all 70 CS ideas

Where Pakistani data lives

• Pakistan Bureau of Statistics: prices (SPI, CPI), census, PSLM surveys and trade
• SBP EasyData: remittances, exchange rates, interest rates and banking
• PSX Data Portal: index and share prices
• opendata.com.pk and Kaggle: scraped datasets for cars, property and e-commerce
• NASA POWER: daily weather for any point in Pakistan
• Cricsheet: ball-by-ball PSL data

15 data science FYP ideas

1. Which cities feel inflation most (Medium)

Turns the PBS weekly Sensitive Price Indicator reports (51 items, 17 cities) into a clean dataset, then compares cities: which items drive each city’s price changes, and where the lowest income groups are hit hardest.

• Data: PBS weekly SPI reports (PDF)
• Cleaning: high, because the tables sit inside PDFs
• Method: index decomposition by item, then clustering cities by their price patterns
• Deliverable: a dashboard with a city comparison

2. Remittances, Eid and the exchange rate (Easy)

Analyses monthly workers’ remittances by country and measures how Eid, Ramazan and exchange-rate moves affect inflows.

• Data: SBP EasyData remittances by country, and exchange rates
• Cleaning: low; you map Gregorian months to the Islamic calendar
• Method: regression with seasonal and Eid effects
• Deliverable: a report with interactive charts

3. Census 2023 district explorer (Easy)

An interactive map of every district showing population growth, density, age structure and urban share from the 2023 census, compared with 2017.

• Data: PBS 2023 census detailed tables (Excel) and the 2017 census
• Cleaning: medium, because some districts were split or renamed between the two censuses
• Method: descriptive statistics, then clustering districts by profile
• Deliverable: a web map

4. How fast used cars lose value (Easy)

Builds depreciation curves by make and model from used car listings, and shows which cars hold their value best in each city.

• Data: PakWheels-based listings on opendata.com.pk and Kaggle
• Cleaning: medium: duplicates and typos in model names
• Method: regression of price on age and mileage for each model, then comparing the curves
• Deliverable: a model comparison web page

5. Property price-per-marla map (Medium)

Maps price per marla across neighbourhoods of Lahore, Karachi and Islamabad, and finds areas priced below what their features suggest.

• Data: Zameen.com property data on Kaggle
• Cleaning: medium: marla, kanal and square feet are mixed, and these are asking prices, not sale prices
• Method: hedonic price regression and spatial clustering
• Deliverable: a map dashboard

6. How the market reacts to SBP rate decisions (Medium)

An event study of the KSE-100 and sector returns around each State Bank policy rate announcement: which sectors gain or lose, and how quickly.

• Data: policy rate history from SBP, and index and share prices from the PSX Data Portal
• Cleaning: low to medium
• Method: an event study with abnormal returns, not price forecasting
• Deliverable: a report and a sector heatmap

7. Heatwave trends in Pakistani cities (Easy)

Measures how the number of extreme-heat days has changed since 1981 in 20 cities, and which cities are warming fastest at night.

• Data: NASA POWER daily temperature and humidity (free API)
• Cleaning: low
• Method: trend tests (Mann-Kendall) and a heat index
• Deliverable: a city comparison dashboard

8. The 2022 floods: who was hit hardest (Hard)

Combines flood extent maps and district damage reports to show where damage was worst relative to population and poverty.

• Data: UNOSAT flood extent layers on HDX (search for “Pakistan floods 2022”), Punjab PDMA daily flood reports (PDF), the 2023 census and PSLM poverty indicators
• Cleaning: high: GIS layers plus PDF reports
• Method: spatial joins, then a regression of damage on exposure and poverty
• Deliverable: a GIS dashboard

9. Out-of-school children by district (Medium)

Maps out-of-school children aged 5 to 16 by district and tests which factors explain the gaps: household income, distance to school, girls’ schools and household size.

• Data: PSLM district-level survey 2019-20 (PBS microdata)
• Cleaning: medium: you must apply the survey weights correctly
• Method: weighted estimates and logistic regression
• Deliverable: a policy brief and a map

10. PSL player value model (Medium)

Rates every Pakistan Super League player by runs and wickets above average for the match situation (powerplay, middle overs, death overs), and suggests a balanced squad within a budget.

• Data: Cricsheet ball-by-ball PSL data
• Cleaning: low
• Method: expected-runs models and knapsack optimisation for squad picks
• Deliverable: player cards and a draft tool

11. Why cash-on-delivery orders get cancelled (Medium)

Finds which orders get cancelled or returned, by product, payment method, discount and timing, and estimates what that costs a seller.

• Data: Pakistan’s Largest E-commerce Dataset (about half a million orders)
• Cleaning: medium
• Method: classification with class imbalance, plus cohort analysis
• Deliverable: a dashboard and recommendations for sellers

12. Export opportunity finder (Medium)

Lists products that Pakistan already exports a little of, where partner countries import a lot and their imports are growing.

• Data: UN Comtrade (free tier) and PBS trade statistics
• Cleaning: medium: HS product codes
• Method: revealed comparative advantage and market growth scoring
• Deliverable: an opportunity table by product and country

13. Wheat yield and weather by district (Hard)

Explains how March temperatures and season rainfall affected wheat yields in Punjab’s districts over 15 or more years.

• Data: Crop Reporting Service Punjab district estimates (yearly PDFs) and NASA POWER weather
• Cleaning: high
• Method: panel regression with district fixed effects
• Deliverable: a report and a yield-risk map

14. The rise of digital payments (Medium)

Builds a clean dataset from SBP’s quarterly Payment Systems Reviews and tracks how Raast, mobile banking and cash withdrawals have changed each quarter.

• Data: SBP Payment Systems Review (quarterly PDFs)
• Cleaning: high
• Method: trend analysis and simple forecasts
• Deliverable: a dashboard

15. What’s really in the petrol price (Medium)

Splits each petrol price revision into the oil price, the exchange rate, and taxes and margins, so you can see what drove each change.

• Data: official price notifications (you compile them into a table), international oil prices and SBP exchange rates
• Cleaning: high, because the notifications are not one table
• Method: price decomposition and regression
• Deliverable: an interactive chart

Frequently asked questions

Does a data science FYP need machine learning?

Not always. A clean new dataset, sound statistics and a clear finding make a strong FYP. Add a model where it answers the question better.

Is Kaggle data acceptable for an FYP?

Yes, if you say where it came from and what its limits are. Scraped listings show asking prices, not sale prices, for example. Combining Kaggle data with an official source makes the project stronger.

How do we get tables out of PDFs?

Use Camelot or Tabula, then check a sample by hand. Document every cleaning step, because panels ask about it.

Next steps

Picked an idea? Write it up with our FYP proposal format and the SRS template for FYP, find a name in 60 FYP title examples, or filter more ideas with the FYP idea generator.

Want to see how data teams work before your build starts? See the Ezitech data science internship and data engineer vs data scientist.