Hacktoberfest: Added code for PCA from scratch under dimensionality-reduction folder

dimensionality-reduction/__init__.py

@@ -0,0 +1 @@
+from pca import pca

dimensionality-reduction/pca.py

@@ -0,0 +1,56 @@
+import numpy as np
+from numpy import linalg as la
+
+
+class pca():
+
+    """
+    PCA is a dimendionality reduction technique.
+    Linear dimensionality reduction using Singular Value Decomposition of the data to project it to a lower dimensional space
+
+    """
+
+    def __init__(self):
+        self.cov=cov
+        self.eigen=eigen
+        self.eigval=None
+        self.eigvct=None
+        self.covar=None
+        self.princicomp=None
+
+    def cov(self,x):
+
+        """
+        first we have to find the covariance matrix of the given array/dataframe
+
+        """
+        self.covar=np.cov(x)
+
+        return self.covar
+
+    def eigen(self,x):
+
+        """
+        Second step is to find eigen values and vectors of the given array  
+
+        """
+        self.eigval, self.eigvct = np.la.eig(self.covar)   
+
+        return self.eigval,self.eigvct
+
+    def ncomp(self,numcomponents):
+
+        """
+        Method to choose number of principal components
+
+        """
+        self.princicomp=self.eigvct[0:numcomponents]
+
+        """
+        For examples, if numcomponents=2 then this function returns first 2 rows of eigenvectors
+
+        """
+
+        return self.princicomp
+
+

requirements.txt

@@ -0,0 +1,5 @@
+numpy
+matplotlib
+pandas
+collections
+os