You are on page 1of 1

Neural Network

Artificial Neural Networks holds utmost importance in data mining techniques. It mimics the way a
human can predict a certain event or value, thus can be a reliable source in formulating models for
forecasting. What it simply requires is ‘data, the more meaningful it is, higher will be the learning it and
more positive outcomes can be achieved.

If dataset having inputs for instance In(x) are provided to neural network, where n is any positive integer,
and y(x) is the output, then neural will try to learn the relationship among them. It uses several
algorithms to achieve the same, however the most used is the backpropagation Neural Networks. It
creates and arbitrary with every input, and try to calculate output. The difference between calculated
output and actual output is then back-propagated and news thus new weights are assigned. Thus it
becomes an iterative procedure where ANN tries to sort out best combination of weights to get close to
actual output. These weights are linear combination of input variables present.

There are two types of neural networks, the ones which are which doesn’t depend on time or simply
they have no time series being presented as input variables are static neural networks, whilst models
that require time series as inputs are dynamic neural networks.

Neural Networks first trains itself to get familiarize with and sort out obscure patterns present among
the dataset. After training, it tests and validates its performance with the similar data. The performance
is evaluated by many KPIs like mean squared error, correlation etc.

Pattern Matching

Pattern matching is a comparison based matching technique which compares the relation between
input and output and try to predict the future outcomes. It usually involves time series data.

Multiple Regression1

Multiple regression is an extension of simple linear regression. It is used when we want to predict the
value of a variable based on the value of two or more other variables. The variable we want to predict is
called the dependent variable (or sometimes, the outcome, target or criterion variable). The variables
we are using to predict the value of the dependent variable are called the independent variables (or
sometimes, the predictor, explanatory or regressor variables).

For example, multiple regression can be used to predict demand by providing historical data having
attributes like temperatures, wind speed, humidity and previous demand.

Multiple regression also allows you to determine the overall fit (variance explained) of the model and
the relative contribution of each of the predictors to the total variance explained. For example, you
might want to know how much of the variation in system demand can be explained by temperatures,
wind speed, humidity and previous demand "as a whole", but also the "relative contribution" of each
independent variable in explaining the variance.

Weights in neural networks are analogous here to regression co-efficients, which are the linear
combination of input variables.

1
https://statistics.laerd.com/spss-tutorials/multiple-regression-using-spss-statistics.php

You might also like