[data analysis]
In data analysis and predictive modeling, a data transformation method that reduces the dimensions of large datasets by transforming large sets of variables into a smaller one that still contains most of the information from the larger sets. This method compresses data by eliminating redundancy and assists in separating features within the data.
[remote sensing]
In remote sensing, a data transformation method that rotates the axes of the input bands to a new multivariate attribute space in which the axes are not correlated. The first component accounts for the greatest variability in the data, the second component accounts for the next largest amount of variability, and so on.