There are several major stages in how Gemval system works:
We continuously add new online shops to our tracking list, so our database of gemstone pricing is steadily growing. To optimize accuracy, we make sure that only reasonable offers are entered into the database. All specimens with noticeably unusual prices are rejected and therefore have no impact upon the prediction model.
We are using supervised machine learning to predict gemstone values based on the specific factors influencing gem pricing – weight, color, clarity, cut and shape. The system can generalize what it learned from the collected data, and it uses this information to predict values for new input data provided by the user.
Apart from providing price prediction based on the trained model, we also analyze and publish useful insights into current gemstone markets such as weight, clarity and price distribution, for example. This information helps gemstone professionals and amateurs to educate themselves about the market and make better-informed decisions concerning trading and collecting.