A prediction robot or bot is a computer software application used to collect and analyze various publicly available or sometimes privately collected or generated data. The main goal is to make future predictions by analyzing obtained information by different data mining methods such as decision trees, clustering and pattern mining. Compared to a human data analyst, a prediction bot can go through vast amounts of data and thus make more accurate predictions than it is possible for a human.
Information is collected through various means, depending on the application and anticipated outcome. Prediction bots can use web spidering (although it is not the best way to obtain the data, because this approach may lead you to “bad” data and you have to check it manually to avoid mistakes), public databases and also information provided directly, such as high-quality data from commercial databases. As the amount of data collected is huge, lots of server space and bandwidth is usually required for building successful applications. It is also noteworthy that web scraping has been discouraged by several large websites such as YAHOO finance, so more and more prediction bots use different API-s and public databases instead.
Analyzing the information is only the beginning and usually the easiest part of making future predictions. The main question is how to analyze the collected data to get best results. The hardest part is reaching a decision – which outcome to predict. Mostly, prediction robots are used as binary classifiers or predictors – they give a yes/no answer. This implementation is the simplest and requires the least resources. Most of the time, prediction bots are not used alone, but a large system is built, consisting of thousands of robots. Each robot is given its own algorithm to start with and as they analyze the data, bots augment their algorithms through machine learning. Final decision is usually a consensus of all the robots, similar to a random forest algorithm.
Prediction robots have a multitude of uses. A good example has been the Web-bot, which crawls the internet and tries to predict various high-importance events, like major stock market crashes and catastrophes. Recently, sports betting has also made use of these, turning them into soccer prediction bots. Prediction robots have also been used in analyzing stock exchanges and FOREX market as there are seemingly endless variables for a human analyst.