We invite you to explore quantum computing and machine learning by tackling a challenge to develop a new approach to wildfire risk and insurance premium modeling in California.
Participants will learn to develop and test on leading platforms in the quantum space. In addition to an attractive prize pool, the finalists will pitch their solution in front of an expert judge panel consisting of executives across finance, quantum, and sustainability.
#quantumcomputing #QuantumML #machinelearning #Wildfireprediction #SustainabilityChallenge
Quantum Machine Learning for Wildfire Prediction
Join Deloitte’s Quantum Sustainability Challenge 2026!
The Competition aims to explore the potential for quantum computing—specifically machine learning algorithms—to enhance wildfire risk prediction and improve insurance premium modeling for urban areas in California facing growing wildfire risks. As the unpredictability and severity of wildfires rises, societies across the globe need to increasingly adapt to the profound financial impacts of these events on local and regional economies.
Advances in modeling and computation technologies can improve the prediction of wildfire risk. Small enhancements in the capability to predict wildfires using complex data which measures wildfire risk using many data sources can save lives, wildlife, and property. Specifically, Quantum Computing offers the potential to improve wildfire detection through Quantum Machine Learning (QML) algorithms. QML algorithms can improve insurance modeling in wildfire-prone regions, improving both private and state-backed insurance programs to better serve the citizens and enterprises they represent.
Specifically, the Competition presents participants with wildfire and insurance datasets for California zip-codes and asks participants to apply quantum technologies to these datasets to tackle a two-fold problem:
Task 1A: Create a quantum algorithm, a (hybrid) quantum machine learning model, that predicts the future risk of a wildfire occurring in California zip-codes in 2023, based on historical data (2018-2022) which are provided in the wildfire dataset. To reduce complexity, a ‘wildfire’ will be defined by if it burns in a wildland setting and is unplanned and uncontrolled. Run your algorithm on a quantum computer or simulator and provide information on the resource requirements of your solution.
Task 1B: Evaluate your solution, describing the advantages and disadvantages of your approach(es). Evaluate the performance differences between your solution and classical approaches.
Task 2: Create a time series model to predict future insurance premiums in 2021 based on historical data (2018-2020) provided in the insurance dataset. Wildfire risk for each zip code is provided by the model output in Task 1 or use the existing fire risk score provided in the dataset.
Supplemental reference for the feature description files — code lookups, column details, additional columns, and data quality notes
These fields use numeric codes from the CAL FIRE FRAP Fire Perimeters dataset. Source: official coded-value domains from the CAL FIRE ArcGIS FeatureServer.
The following clarifications map the feature description files to the actual dataset columns.
These columns are present in the data and not covered in the feature description files.
year_month column
Some rows contain full dates (YYYY-MM-DD) or leaked ZIP code values instead of the expected YYYY-MM format. Clean this column before use.
OBJECTIVE column
Approximately 20 rows contain date strings instead of the expected 1 or 2 numeric codes. This is a processing artifact
— treat these as missing values.
AGENCY_ID encoding
Missing AGENCY values are encoded as integer 9 rather than NaN. This is indistinguishable from a valid category code unless you are aware of it. If using AGENCY_ID, consider cross-referencing with the original AGENCY column.
Double-space column names
Some insurance column names contain embedded double-spaces (e.g., "CAT Cov A Fire - Incurred Losses") due to newline removal during preprocessing. Use care when referencing these columns by name.
Detailed Challenge Briefing: Background & Tasks
You can review the full Challenge Description via this link.
Task 1 Data
Please use these data sets to complete task 1:
Task 2 Data
Please use these data sets to complete task 2:
Once participants have registered for the challenge, they can develop their solution to the challenge using either of the quantum development platforms below provided by IBM and AWS:
About the IBM Quantum Platform and Qiskit:
About AWS Braket: