Stratohack – Final Problem Statements (with Bonus Points)
Hackathon Kickoff: Choose Your Challenge!
Welcome, innovators!
We’ve lined up a series of exciting AI/ML problem statements for you. Your mission? Go
through the list, pick the one that excites you the most, and dive in!
Each statement comes with unique challenges and bonus opportunities—so keep an eye
out for those extra points that can set your team apart. Remember, it’s not just about
solving the problem; it’s about showing creativity, technical skill, and out-of-the-box
thinking.
Pick your challenge, start building, and let the hackathon magic begin!
Pro Tips for Choosing Your Statement:
● Look for a problem that matches your team’s strengths.
● Check for bonus points—you can get an edge with some extra effort.
● Choose something that inspires you; passion fuels better solutions!
● Don’t overthink—start early, iterate quickly, and showcase your creativity.
1. Predict Aircraft Engine Remaining Useful Life (RUL)
Problem: Predict the remaining useful life of an aircraft engine using multi-sensor
time-series data (temperature, vibration, pressure, speed).
Dataset: NASA C-MAPSS Turbofan Engine Dataset.
How to Approach:
● Baseline: Extract features (avg, slope, variance) → Random Forest/XGBoost.
● Advanced: Train LSTM/GRU on time-series sequences.
Bonus Points:
● Visualize engine degradation over time.
● Deploy a dashboard that predicts RUL in real-time as new sensor data streams.
● Add anomaly detection to catch sudden failures.
2. Dynamic Airline Ticket Price Forecasting
Problem: Predict airline ticket prices by considering not only routes but also fuel prices,
holidays, and seasonality.
Dataset: Kaggle Flight Price + IATA Fuel Index + Holiday Calendar.
How to Approach:
● Baseline: Regression with tabular flight features.
● Advanced: Time-series + external features (Prophet, Transformers).
Bonus Points:
● Build a "Price Alert Bot" that notifies cheapest time to fly.
● Simulate impact of changing fuel prices on ticket cost.
● Predict ticket price uncertainty (confidence intervals).
3. Multi-label Cloud Classification & Segmentation
Problem: Classify multiple cloud types in the same image and segment cloudy regions in
sky/satellite images.
Dataset: NASA/MODIS Cloud Imagery.
How to Approach:
● Baseline: CNN with transfer learning.
● Advanced: U-Net / Mask R-CNN for segmentation.
Bonus Points:
● Build a “Cloud Map” of the sky using segmentation masks.
● Predict solar power generation impact from cloud cover.
● Add explainability: highlight which parts of image influenced prediction.
4. Flight Delay Duration Prediction with Weather Integration
Problem: Predict delay duration (minutes) using flight data + real weather data.
Dataset: Kaggle Flight Delay + NOAA Weather Data.
How to Approach:
● Baseline: XGBoost regression.
● Advanced: Weather + congestion features, feature importance with SHAP.
Bonus Points:
● Build a live demo: input flight + weather → predicted delay.
● Add explainability charts: “Weather caused 60% of this delay.”
● Forecast total delay hours for an airport/day.
5. Airport Traffic Flow Prediction (Graph ML)
Problem: Model global airport connectivity and predict high-congestion airports using
graph-based ML.
Dataset: OpenSky ADS-B Flight Data.
How to Approach:
● Baseline: Clustering on airport traffic.
● Advanced: Graph Neural Networks for prediction.
Bonus Points:
● Visualize airport network with interactive graph.
● Predict cascading delays: if JFK is congested, which airports are next?
● Add a simulation of rerouting strategies.
6. Satellite Image Change Detection (Earth Observation)
Problem: Detect and classify changes (urbanization, deforestation, flooding, etc.) in
Earth’s surface using multi-temporal satellite images.
Dataset: Sentinel-2 / Landsat Multi-temporal Imagery (available on Google Earth
Engine).
How to Approach:
● Baseline: Preprocess → CNN classifier for binary “change/no-change.”
● Advanced: Siamese Networks / Transformers for pixel-wise temporal change
detection.
Bonus Points:
● Build an interactive “Change Heatmap” of any chosen region.
● Add explainability: highlight which regions caused the most change.
● Forecast future land cover change trends using time-series ML.
7. Space Weather Forecasting (Solar Storm Prediction)
Problem: Predict solar flares or coronal mass ejections (CMEs) that affect satellites,
astronauts, and communication systems.
Dataset: NASA Solar Dynamics Observatory (SDO) + NOAA Space Weather Prediction
Center Data.
How to Approach:
● Baseline: Feature engineering on solar activity indices → XGBoost/Random Forest.
● Advanced: LSTM/Transformer models on solar magnetogram time-series.
Bonus Points:
● Deploy a real-time solar storm alert dashboard.
● Visualize solar flare probability with uncertainty bands.
● Simulate impact on GPS / satellite communication outage.
8. Astronaut Health Monitoring with AI
Problem: Predict astronaut health anomalies (stress, fatigue, radiation impact) from
wearable sensor + mission data.
Dataset: NASA Astronaut Physiology Data (synthetic datasets available).
How to Approach:
● Baseline: Tabular sensor data → anomaly detection models.
● Advanced: Multi-modal ML (physiological + environmental + mission logs).
Bonus Points:
● Build a personalized astronaut health dashboard.
● Predict long-term health risk (bone loss, cardiovascular issues).
● Add voice/text log analysis (sentiment/emotion recognition from astronaut logs).
9. Space Debris Collision Risk Prediction
Problem: Predict the probability of collision between satellites and space debris using
orbital trajectory data.
Dataset: CelesTrak / NORAD TLE + ESA Space Debris Database.
How to Approach:
● Baseline: Rule-based proximity analysis with TLE data.
● Advanced: Graph Neural Networks or Reinforcement Learning for dynamic collision
avoidance.
Bonus Points:
● Build a real-time orbital visualization of debris threats.
● Suggest collision-avoidance maneuvers (Delta-V optimization).
● Rank satellites by “risk score” over time.
10. Predict Rocket Launch Success Probability¶
Problem: Predict whether a rocket launch will succeed or fail based on weather, rocket
specs, and historical launch data.
Dataset: Kaggle Rocket Launch Data + NOAA Weather Data.
How to Approach:¶
● Baseline: Logistic Regression / Random Forest classification.¶
● Advanced: Ensemble models with environmental + technical features.¶
Bonus Points:¶
● Build a launch readiness prediction tool (dashboard/API).¶
● Add explainability: “Weather caused 70% of failure probability.”¶
● Simulate future launch reliability trends per company/rocket.¶
Problem Statement Dataset / Source Link(s) / Notes
1. Predict Aircraft Engine NASA C-MAPSS NASA’s CMAPSS Jet Engine Simulated Data:
Remaining Useful Life (Commercial Modular NASA Open Data Portal — CMAPSS Jet Engine
(RUL) Aero-Propulsion System Data NASA Data
Simulation) Direct zip: [Link] NASA Data
Also in prognostics repository: NASA
Prognostics Health Management (P&H)
Repository NASA+2GitHub+2
2. Dynamic Airline Ticket Several Kaggle datasets Examples: “Flight Prices” dataset on Kaggle
Price Forecasting of “Flight Price / Flight (Expedia tickets etc.) Kaggle
Fare / Ticket Price” “Flight Fare Prediction – Ease My Trip dataset”
Kaggle+1
3. Multi-label Cloud MODIS Cloud Products; MODIS Cloud Product (MOD06 / MOD06_L2
Classification & Sentinel‐2 / Landsat etc.) – physical and radiative cloud properties.
Segmentation imagery; specialized MODIS+1
cloud classification MODIS products via Earth Engine & LAADS
datasets etc. National Snow and Ice Data
Center+2MODIS+2
Cumulo: cloud labels merged with CloudSat +
MODIS for cloud classification tasks. arXiv
SEN12MS: Sentinel‐1/2, MODIS land cover
maps etc. (useful for segmentation/land cover
tasks) arXiv
4. Flight Delay Duration Flight Delay datasets + Specific flight delay data: e.g., some Kaggle
Prediction with Weather Weather datasets “Flight Delay” datasets; can also pull from
Integration (NOAA, etc.) government aviation authorities.
NOAA / National Weather Services for historical
weather. (You’ll need to join by time & location.)
(No single perfect public unified dataset found in
search for exact combination, but multiple sources
available.)
5. Airport Traffic Flow OpenSky Network OpenSky Network: historical and live ADS-B
Prediction (Graph ML) ADS-B / Mode S / data. OpenSky Network+1
historical flight For flight schedules / routings, there are public
trajectory data datasets (IATA, OAG etc.) though licensing may
vary.
6. Satellite Image Change Sentinel-2, Landsat, Sentinel-2 / Landsat imagery via USGS / ESA /
Detection (EO) MODIS, Google Earth Copernicus open data portals
Engine MODIS (various land cover / surface reflectance
products) via NASA / LAADS / Earthdata (Earth
Engine)
SEN12MS as multi-sensor example. arXiv
7. Space Weather Solar Dynamics e.g., NOAA SWPC data on solar flares,
Forecasting (Solar Storm Observatory (SDO), geomagnetic indices; SDO magnetogram / solar
Prediction) NOAA Space Weather imagery datasets from NASA. (You’ll likely pull
Prediction Center, etc. from NASA/NOAA websites.)
8. Astronaut Health NASA astronaut NASA has some astronaut health / physiology
Monitoring with AI physiology datasets; datasets (public ones may be limited) + synthetic
mission logs; datasets; also research articles often provide
simulated/synthetic data processed / synthetic data. (Might need to
synthesize or request via NASA/academic
partners.)
9. Space Debris Collision CelesTrak / NORAD CelesTrak TLE data for satellite orbits.
Risk Prediction TLE (Two Line Element ESA’s space debris / orbital conjunction data.
sets) + ESA / Space Combined with ground station / orbit
Debris databases propagation tools.
10. Predict Rocket Launch Historical launch records Datasets of rocket launches (success/failure) e.g.
Success Probability + weather archives via SpaceX / Russian / NASA launch logs;
combined with weather data from NOAA etc.
Some Kaggle datasets may have launch data; also
archival launch databases (e.g. Gunter's Space
Page, Space-Track).