Machine Learning

Overview
Machine Learning is the processing of crucial data with the use of algorithms on data structures. The algorithms are then analyzed and converted into formulas and patterns adaptable for economic and census use. Therefore, the methodology itself is machine learning. Internet technology uses different methods to capture data such as HTTP Cookie and data tracking but once analyzed machine learning can be used to project statistical data.

Predictive or supervised learning
Predictive or supervised learning operates in a manner of functions. The relationship between x and output y is mapped using features, attributes, or covariates. However; even though mathematical and textual patterns are recognized relationships between objects such as images, sentences, context, and graphs are also interpreted. As such y can be considered categorical or nominal. The relationship between x and y can then be examined as pattern recognition, regressive, or ordinal regression. Examples include filtering, email spam, and image recognition.

Descriptive or unsupervised learning
Descriptive or unsupervised learning uses one single set of data such as x and finds a significant pattern among that data. Because the data is without an output it is usually harder to find a recognizable pattern. Patterns within the data are malleable due to the process itself not having ample evidence (the lack of an output y). The behavior of the data examined can only be related through the clustering of present data values. An example of unsupervised learning would be page ranking algorithms.

Reinforcement learning
Reinforcement learning is used when data has a third set directly correlated to y. It can be interpreted as reward or punishment signals. This type of machine learning operates based on the psychological behaviors of users and the effect brought on by exposure to an environment. Popular examples would be game theory or economics.

Problems with Machine Learning
Software and technology can only be taught to recognize patterns already archived by its creator and its own limitations. This type of learning is subjected to the current bounds of technology and leaves no room for self-adaption. Without the proper protocols, machine learning can be rendered useless due to the random nature of human interactive data. If successful machine learning can be used to violate the privacy of unsuspecting users through the patterns they detect.