Tutorial

This example shows how saiph works.

import pandas as pd

import saiph
from saiph.visualization import plot_circle

Fit the model

df = pd.read_csv("../../../tests/fixtures/iris.csv")

coord, model = saiph.fit_transform(df, nf=5)
print(coord.head())

     Dim. 1    Dim. 2    Dim. 3    Dim. 4    Dim. 5
0 -2.643907 -0.603966 -0.062524  0.123714  0.023531
1 -2.467082  0.160855  0.760712  0.380099 -0.121408
2 -2.715591 -0.064140  0.572116  0.034530 -0.040782
3 -2.654857  0.102214  0.759613  0.026693 -0.025427
4 -2.753830 -0.717357 -0.158470 -0.050660  0.072761

Project individuals

saiph.visualization.plot_projections(model, df, (0, 1))

Get statistics about the projection

model = saiph.stats(model, df)
print(model.cos2)

                Dim. 1    Dim. 2        Dim. 3        Dim. 4        Dim. 5
sepal.length  0.558969  0.005336  9.419276e-03  6.589394e-03  1.118242e-07
sepal.width   0.055004  0.258351  5.514559e-02  4.731168e-04  3.677468e-07
petal.length  0.968523  0.000002  3.960561e-06  2.223127e-07  4.835432e-06
petal.width   0.882583  0.000174  7.936563e-10  1.106446e-03  1.948436e-04
variety       0.465027  0.278545  3.323229e-02  1.535270e-04  4.003553e-05

Correlation circle

plot_circle(model=model)

Variable contributions

print(model.contributions)

                 Dim. 1     Dim. 2     Dim. 3     Dim. 4     Dim. 5
sepal.length  19.318137   5.442424  16.402147  52.632696   1.256580
sepal.width    6.059927  37.868647  39.686917  14.103181   2.278750
petal.length  25.428824   0.098688   0.336334   0.305714   8.263041
petal.width   24.274426   0.982162   0.004761  21.567410  52.452391
variety       24.918686  55.608080  43.569841  11.391000  35.749238

---

saiph.visualization.plot_var_contribution(
    model.contributions["Dim. 1"].to_numpy(), model.contributions.index.to_numpy()
)

Explained variance

print(model.explained_var)

[1.73161455e-04 6.00547784e-05 2.64746674e-05 6.90060129e-06
 1.18971992e-06]

---

print(model.explained_var_ratio)

[0.64665272 0.22426807 0.09886678 0.02576955 0.00444288]

---

saiph.visualization.plot_explained_var(model)