Implementing Machine Learning Methods in Searching Processes

Methods of machine learning are currently very widespread, popular and are used in number of sectors - whether it is medicine, industry or the transportation. In many industries, machine learning is a factor of improvement which streamlines the process of disease diagnosing, speeds up the process of object identification at airports or eliminates number of errors which may occur in the production process based on previous testing. Machine learning applied to data retrieval processes in various types of databases, whether relational or non-relational ones, can bring more benefits than minimization of data retrieval time or reduction of database server usage. Based on the previous research - which focused on a comparison of the time required to obtain data in relational and non-relational databases - we concluded that it is more appropriate to implement methods and processes of machine learning to non-relational key-value type databases such as MongoDB or DynamoDB. Our proposed solution works with two principles. The first one is the principle of monitoring unfinished commands (operations) and their subsequent transfer to the buffer memory. The second principle is based on definition of the limit at which can machine learning efficiently provide appropriate transfer of the supposedly requested data to the buffer. This action can not only speed up the time required to obtain data, but also provide proposal of data selection operations based on previous queries of the user.