How It Works
Raqam uses advanced machine learning algorithms trained on publicly available historical data along with crowdsourced input collected through our reported valuations to predict prices.
The models are robust as they actively learn from multiple sources while using intelligent feature enhancements that give it an edge over generic models. User reported valuations are used for reinforcement learning.
We retrain frequently to maintain accuracy. Raqam uses confidence pills to measure it's own confidence while making an estimation.
Raqam has been trained over 200k+ rows with 6 seperate machine learning models for 4 digit, 5 digit, 6 digit along with 3 car valuation models. Every prediction comes with a confidence percentage declared by the underlying model itself.
Based on the density of the training data, certain predictions are more confident and accurate than others as mentioned in the FAQs. On top of traditional machine learning, Raqam uses over 200+ weighted parameters to detect different patterns in any given number.
Raqam uses over 10 parameters to train it's car valuation models. The weights of these patterns (which in turn determines the importance of a pattern in the market) has been choosen from the training data itself.
Our ML models are fully trained and built in Qatar to serve within Qatar. For further information, please contact us directly.