Using machine learning to determine whether network components are being used as intended

In some examples, a server may receive, from a software agent, data associated with a particular component of a plurality of components in a system (e.g., an information technology (IT) network). The data may include connection data, network location data, software data, user data, hardware data, and network routing data. The server may determine, using a k-nearest classification algorithm and based on the data, a current usage of the particular component. The server may determine an intended usage of the particular component and perform a comparison of the current usage with the intended usage. If the server determines that the current usage differs from the intended usage by at least a predetermined percentage, then the server may perform one or more remediation actions to modify the current usage to differ from the intended usage by less than the predetermined percentage.