fancyimpute

0.7.0 verified Fri May 01 auth: no python maintenance

Matrix completion and feature imputation algorithms for Python. Current version is 0.7.0 (last release 2020). The library provides iterative imputation methods like SoftImpute, IterativeImputer, KNN, and NuclearNormMinimization. It is in maintenance mode with no recent updates.

pip install fancyimpute

Common errors

error ImportError: No module named 'fancyimpute' ↓

cause fancyimpute not installed or installed in wrong environment.

fix

Run: pip install fancyimpute

error ValueError: Input contains NaN, infinity or a value too large for dtype('float64') ↓

cause Data contains infinity or extremely large values.

fix

Replace infinity with NaN using np.isinf, or scale data.

error ModuleNotFoundError: No module named 'cvxpy' ↓

cause Attempting to use NuclearNormMinimization without cvxpy installed.

fix

Install cvxpy: pip install cvxpy

error AttributeError: 'SoftImpute' object has no attribute 'fit_transform' ↓

cause Old version of fancyimpute (<0.5.0) where fit() returned imputed matrix.

fix

Upgrade fancyimpute: pip install --upgrade fancyimpute

Warnings

deprecated fancyimpute is no longer actively maintained (last release 2020). Consider using sklearn.impute.IterativeImputer for iterative imputation or other modern alternatives. ↓

fix Use sklearn.impute.IterativeImputer or other up-to-date libraries.

gotcha Input data must be a 2D numpy array with NaN values for missing entries. If you use a pandas DataFrame, it will be converted but columns may be reordered. ↓

fix Ensure data is a 2D numpy array with NaN for missing values. Convert pandas DataFrame with .values.

gotcha Some solvers (e.g., NuclearNormMinimization) require cvxpy, which is an optional dependency. If not installed, import will fail. ↓

fix Install cvxpy: pip install cvxpy

breaking In version 0.5.0, the API changed: fit() returns a fitted model, and transform() or fit_transform() must be used to impute. Older versions used fit() to return imputed data. ↓

fix Use fit_transform() or fit() then transform() instead of relying on fit() returning imputed data.

Imports

SimpleFill
```
from fancyimpute import SimpleFill
```
correct import
KNN
```
from fancyimpute import KNN
```
correct import
SoftImpute
```
from fancyimpute import SoftImpute
```
correct import

IterativeImputer

wrong

from fancyimpute.iterative_imputer import IterativeImputer

correct

from fancyimpute import IterativeImputer

wrong: internal submodule not exposed

NuclearNormMinimization

from fancyimpute import NuclearNormMinimization

correct import

Quickstart

Basic usage: impute missing values using SoftImpute.

import numpy as np
from fancyimpute import SoftImpute

# Create data with missing values
X = np.array([[1, 2, np.nan], [4, np.nan, 6], [7, 8, 9]])
# impute using SoftImpute
X_filled = SoftImpute().fit_transform(X)
print(X_filled)