Insurance Predictive Models
Unique analytical competency to combine insurance specialized predictive modelling with modern machine-learning algorithms.
Full Coverage analytics on every single step in the insurance processes.
* Churn Rate Analysis - Prediction accuracy +11% vs. benchmark results of other participants.
* Conversion Model - Be rewareded on monthly performance uplift.
* Claim Reserves Prediction - Proven to respect more prudent reserving philosophy.
* Vehicle Group Classifiction - Verified by client company's own multi-year efforts.
* Anti-fraud Algorithms - More robust proprietary ML algorithms verified by the clients.
IoT Sensor Signal Analysis
Extensive knowledge of IoT sensors, both on firmware and software.
Verified results by clients of telematics related analytics.
* Car Accident Reconstruction - Proprietary algorithms to calculate accident impact point and impact force, results verified to be more precises than device suppliers.
* Car Accident Detection - Proprietary algorithms to detect whether a crash is a real accident.
* Car Accident Damage Estimation - Proprietary algorithms to estimate the damage (both physical damage and body injury) based on telematics crash data.
* Telematics Driving Footprint Competition - Achieve prediction accuracy of 99.99%.
Natural Language Processing
Up to 80% of data stored by organisations is unstructured and a major part of it is in the free text form. However free texts CANNOT be easily used in analytics reports or models.
Built own proprietary insurance-specialized dictionaries and NLP libraries to erform text-mining extracts useful features/intelligence hidden in the free texts to structured format.
* Ultimate Claims Costs Prediction - Text-mining in extracting the claims handling notes.
* Claims Status Evolvement Prediction - Text-mining in extracting the claims handling notes.
* Fraud Detection - Text-mining in building fraud relation graphs.
* Fraud Prevention - Text-mining to extract features for underwriting usage.
Up to 80% of data stored by organisations is unstructured and a significant part of it is in image format of huge values.
The use of Deep-learning methods such as Convolutional Neural Network (CNN) has proven to achieve significant results.
* Claims Damage Estimation - Speed up settlements with accident and inspection photos.
* Medical Imagery - Explore usuage in health insurance.
* Satellite Imagery - Explore usuage in agriculture, property, marine and weather insurance.
* Drone Imagery - Explore usuage in agriculture, property and construction insurance.