How to reduce dimentionality on Sparse Matrix in Python?
How to reduce dimentionality on Sparse Matrix in Python?
Recipe Objective
While working on a large dataset with many feature and after creating a Sparse Matrix and training a model it takes a high computational cost. Managing and vizualizing the matrix is also very difficult. So we need to reduce the dimension of the matrix.
So this recipe is a short example of how can reduce dimentionality on Sparse Matrix in Python.
Step 1 - Import the library - GridSearchCv
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import TruncatedSVD
from scipy.sparse import csr_matrix
from sklearn import datasets
Here we have imported various modules like StandardardScaler, datasets, TruncatedSVD and csr_matrix from differnt libraries. We will understand the use of these later while using it in the in the code snipet.
For now just have a look on these imports.
Step 2 - Setup the Data
Here we have used datasets to load the inbuilt digits dataset. We have used standardscaler to scale the data such that the mean becomes 0 and standard deviation to 1. We have also made a sparse matrix of the data by the function csr_matrix.
digits = datasets.load_digits()
X = StandardScaler().fit_transform(digits.data)
print(X)
X_sparse = csr_matrix(X)
print(X_sparse)
Step 3 - Using GridSearchCV
We can truncate the marix that is we can reduce the dimension of the matrix by using the function TruncatedSVD with a parameter n_components which shows the final number of fetures we want. So we have fit and transform the matrix in the function to get the truncated matrix.
tsvd = TruncatedSVD(n_components=10)
X_sparse_tsvd = tsvd.fit(X_sparse).transform(X_sparse)
print(); print(X_sparse_tsvd)
Step 6 - Printing Results
Now we are using print statements to print the results.
print("Original number of features:", X_sparse.shape[1])
print("Reduced number of features:", X_sparse_tsvd.shape[1])
print(); print(tsvd.explained_variance_ratio_[0:6].sum())
As an output we get:
[[ 0. -0.33501649 -0.04308102 ... -1.14664746 -0.5056698 -0.19600752] [ 0. -0.33501649 -1.09493684 ... 0.54856067 -0.5056698 -0.19600752] [ 0. -0.33501649 -1.09493684 ... 1.56568555 1.6951369 -0.19600752] ... [ 0. -0.33501649 -0.88456568 ... -0.12952258 -0.5056698 -0.19600752] [ 0. -0.33501649 -0.67419451 ... 0.8876023 -0.5056698 -0.19600752] [ 0. -0.33501649 1.00877481 ... 0.8876023 -0.26113572 -0.19600752]] (0, 1) -0.3350164872543856 (0, 2) -0.04308101770538793 (0, 3) 0.2740715207154218 (0, 4) -0.6644775126361527 (0, 5) -0.8441293865949171 (0, 6) -0.40972392088346243 (0, 7) -0.1250229232970408 (0, 8) -0.05907755711884675 (0, 9) -0.6240092623290964 (0, 10) 0.4829744992519545 (0, 11) 0.7596224512649244 (0, 12) -0.05842586308220443 (0, 13) 1.1277211297338117 (0, 14) 0.8795830595483867 (0, 15) -0.13043338063115095 (0, 16) -0.04462507326885248 (0, 17) 0.11144272449970435 (0, 18) 0.8958804382797294 (0, 19) -0.8606663175537699 (0, 20) -1.1496484601880896 (0, 21) 0.5154718747277965 (0, 22) 1.905963466976408 (0, 23) -0.11422184388584329 (0, 24) -0.03337972630405602 (0, 25) 0.48648927722411006 : : (1796, 38) -0.8226945146290309 (1796, 40) -0.061343668908253476 (1796, 41) 0.8105536026095989 (1796, 42) 1.3950951873625397 (1796, 43) -0.19072005925701047 (1796, 44) -0.5868275383619802 (1796, 45) 1.3634658076459107 (1796, 46) 0.5874903313016945 (1796, 47) -0.08874161717060432 (1796, 48) -0.035433262605025426 (1796, 49) 4.179200682513991 (1796, 50) 1.505078217025183 (1796, 51) 0.0881769306516768 (1796, 52) -0.26718796251356636 (1796, 53) 1.2010187221077009 (1796, 54) 0.8692294429227895 (1796, 55) -0.2097851269640334 (1796, 56) -0.023596458909150665 (1796, 57) 0.7715345500122912 (1796, 58) 0.47875261517372414 (1796, 59) -0.020358468129093202 (1796, 60) 0.4441643511677691 (1796, 61) 0.8876022965425754 (1796, 62) -0.26113572420685327 (1796, 63) -0.1960075186604789 [[ 1.91421562 -0.95449937 -3.94604425 ... 1.4963196 0.1160377 -0.80839011] [ 0.58898173 0.9246434 3.92476559 ... 0.55743317 1.08360629 0.07914133] [ 1.30203646 -0.31719139 3.02334129 ... 1.15547162 0.78332798 -1.12203121] ... [ 1.02259528 -0.14791152 2.46997819 ... 0.52912028 2.04799351 -2.0550423 ] [ 1.07605482 -0.38090797 -2.45549106 ... 0.76221796 1.07481616 -0.33991093] [-1.25770756 -2.22760395 0.28362814 ... -1.20258084 0.80783614 -1.84480729]] Original number of features: 64 Reduced number of features: 10 0.4561203224142434
