add Factor Analysis, Truncated SVD, Kernel PCA, MDS and Isomap dimens…

…ionaity reduction techniques
matteo-serafino · Jul 10, 2022 · 40ace20 · 40ace20
1 parent 8970b6e
commit 40ace20
Showing 1 changed file with 172 additions and 0 deletions.
diff --git a/src/dimensionality_reduction.py b/src/dimensionality_reduction.py
@@ -1,7 +1,13 @@
+import imp
 import umap
 import pandas as pd
 from sklearn.manifold import TSNE
+from sklearn.manifold import MDS
+from sklearn.manifold import Isomap
 from sklearn.decomposition import PCA
+from sklearn.decomposition import FactorAnalysis
+from sklearn.decomposition import TruncatedSVD
+from sklearn.decomposition import KernelPCA
 from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
 from sklearn.preprocessing import StandardScaler
 
@@ -136,3 +142,169 @@ def pca(self, X, y, n_components:int = 3):
         Plot.matrix_plot(df=df_pca)
 
         return df_pca
+
+    def factor_analysis(self, X, y, n_components:int = 3):
+
+        X_scaled = StandardScaler().fit(X).transform(X)
+
+        if (n_components == 3):
+
+            print("Factor Analysis 3d dimensionality reduction running...")
+            df_fa = pd.DataFrame(FactorAnalysis(n_components=n_components, rotation="varimax", random_state=0).fit_transform(X_scaled), 
+                columns=["component_1", "component_2", "component_3"])
+            df_fa["label"] = y
+            Plot.plot_3d(df=df_fa, title="Factor Analysis")
+
+        elif(n_components == 2):
+
+            print("Factor Analysis 2d dimensionality reduction running...")
+            df_fa = pd.DataFrame(FactorAnalysis(n_components=n_components, rotation="varimax", random_state=0).fit_transform(X_scaled),
+                columns=["component_1", "component_2"])
+            df_fa["label"] = y
+            Plot.plot_2d(df=df_fa, title="Factor Analysis")
+
+        elif(n_components == 1):
+
+            print("Factor Analysis 1d dimensionality reduction computing...")
+            df_fa = pd.DataFrame(FactorAnalysis(n_components=n_components, rotation="varimax", random_state=0).fit_transform(X_scaled),
+                columns=["component_1"])
+            df_fa["label"] = y
+
+        else:
+            return None
+
+        Plot.matrix_plot(df=df_fa)
+
+        return df_fa
+
+    def truncated_SVD(self, X, y, n_components:int = 3):
+
+        X_scaled = StandardScaler().fit(X).transform(X)
+
+        if (n_components == 3):
+            print("Truncated SVD 3d dimensionality reduction running...")
+            df_tsvd = pd.DataFrame(TruncatedSVD(n_components=n_components, random_state=0).fit_transform(X_scaled), 
+                columns=["component_1", "component_2", "component_3"])
+            df_tsvd["label"] = y
+            Plot.plot_3d(df=df_tsvd, title="Truncated SVD")
+
+        elif(n_components == 2):
+
+            print("Truncated SVD 2d dimensionality reduction running...")
+            df_tsvd = pd.DataFrame(TruncatedSVD(n_components=n_components, random_state=0).fit_transform(X_scaled),
+                columns=["component_1", "component_2"])
+            df_tsvd["label"] = y
+            Plot.plot_2d(df=df_tsvd, title="Truncated SVD")
+
+        elif(n_components == 1):
+
+            print("Truncated SVD 1d dimensionality reduction computing...")
+            df_tsdv = pd.DataFrame(TruncatedSVD(n_components=n_components, random_state=0).fit_transform(X_scaled),
+                columns=["component_1"])
+            df_tsdv["label"] = y
+
+        else:
+            return None
+
+        Plot.matrix_plot(df=df_tsvd)
+
+        return df_tsvd
+
+    def kernel_pca(self, X, y, n_components:int):
+
+        X_scaled = StandardScaler().fit(X).transform(X)
+
+        if (n_components == 3):
+            print("Kernel PCA 3d dimensionality reduction running...")
+            df_kpca = pd.DataFrame(KernelPCA(n_components=n_components, kernel='rbf', gamma=15, random_state=42).fit_transform(X_scaled), 
+                columns=["component_1", "component_2", "component_3"])
+            df_kpca["label"] = y
+            Plot.plot_3d(df=df_kpca, title="Kernel PCA")
+
+        elif(n_components == 2):
+
+            print("Kernel PCA 2d dimensionality reduction running...")
+            df_kpca = pd.DataFrame(KernelPCA(n_components=n_components, kernel='rbf', gamma=15, random_state=42).fit_transform(X_scaled),
+                columns=["component_1", "component_2"])
+            df_kpca["label"] = y
+            Plot.plot_2d(df=df_kpca, title="Kernel PCA")
+
+        elif(n_components == 1):
+
+            print("Kernel PCA 1d dimensionality reduction computing...")
+            df_kpca = pd.DataFrame(KernelPCA(n_components=n_components, kernel='rbf', gamma=15, random_state=42).fit_transform(X_scaled),
+                columns=["component_1"])
+            df_kpca["label"] = y
+
+        else:
+            return None
+
+        Plot.matrix_plot(df=df_kpca)
+
+        return df_kpca
+
+    def multidim_scaling(self, X, y, n_components:int):
+
+        X_scaled = StandardScaler().fit(X).transform(X)
+
+        if (n_components == 3):
+            print("MDS 3d dimensionality reduction running...")
+            df_mds = pd.DataFrame(MDS(n_components=n_components, metric=True, random_state=42).fit_transform(X_scaled), 
+                columns=["component_1", "component_2", "component_3"])
+            df_mds["label"] = y
+            Plot.plot_3d(df=df_mds, title="MDS")
+
+        elif(n_components == 2):
+
+            print("MDS 2d dimensionality reduction running...")
+            df_mds = pd.DataFrame(MDS(n_components=n_components, metric=True, random_state=42).fit_transform(X_scaled),
+                columns=["component_1", "component_2"])
+            df_mds["label"] = y
+            Plot.plot_2d(df=df_mds, title="MDS")
+
+        elif(n_components == 1):
+
+            print("MDS 1d dimensionality reduction computing...")
+            df_mds = pd.DataFrame(MDS(n_components=n_components, metric=True, random_state=42).fit_transform(X_scaled),
+                columns=["component_1"])
+            df_mds["label"] = y
+
+        else:
+            return None
+
+        Plot.matrix_plot(df=df_mds)
+
+        return df_mds
+
+    def isomap(self, X, y, n_components:int):
+
+        X_scaled = StandardScaler().fit(X).transform(X)
+
+        if (n_components == 3):
+            print("Isomap 3d dimensionality reduction running...")
+            df_isomap = pd.DataFrame(Isomap(n_neighbord=10, n_components=n_components).fit_transform(X_scaled), 
+                columns=["component_1", "component_2", "component_3"])
+            df_isomap["label"] = y
+            Plot.plot_3d(df=df_isomap, title="Isomap")
+
+        elif(n_components == 2):
+
+            print("Isomap 2d dimensionality reduction running...")
+            df_isomap = pd.DataFrame(Isomap(n_neighbord=10, n_components=n_components).fit_transform(X_scaled),
+                columns=["component_1", "component_2"])
+            df_isomap["label"] = y
+            Plot.plot_2d(df=df_isomap, title="Isomap")
+
+        elif(n_components == 1):
+
+            print("Isomap 1d dimensionality reduction computing...")
+            df_isomap = pd.DataFrame(Isomap(n_neighbord=10, n_components=n_components).fit_transform(X_scaled),
+                columns=["component_1"])
+            df_isomap["label"] = y
+
+        else:
+            return None
+
+        Plot.matrix_plot(df=df_isomap)
+
+        return df_isomap