How to Remove Rows in Numpy Array that Contains Non-Numeric Values?

Last Updated : 13 Dec, 2025

Given a NumPy array that may contain invalid values such as NaN, inf, or non-numeric entries, the task is to remove all rows that contain any such value and keep only rows with clean numerical data.

For Example:

Input: [[10.5, 22.5, 3.8], [41, nan, nan]]
Output: [[10.5, 22.5, 3.8]]

Let's explore different methods to remove rows which contains non numeric values in Python.

Using np.isfinite()

This method keeps only those rows where every element is a valid finite number. It works by detecting and removing rows that contain NaN, inf, or -inf in one check.

Python 
import numpy as np

a = np.array([[10.5, 22.5, 3.8],
              [41, np.nan, np.nan]])

res = a[np.isfinite(a).all(axis=1)]
print(res)

Output
[[10.5 22.5  3.8]]

Explanation:

  • np.isfinite(a) creates a mask where valid numbers = True, invalid values = False.
  • .all(axis=1) keeps rows where all values are True.
  • a[...] filters and returns only numeric rows

Using ~np.isnan(a).any(axis=1)

This method filters rows by checking if they contain any NaN. Rows without NaN are kept; rows with even one NaN are removed.

Python 
import numpy as np

a = np.array([[10.5, 22.5, 3.8],
              [41, np.nan, np.nan]])

clean = a[~np.isnan(a).any(axis=1)]
print(clean)

Output
[[10.5 22.5  3.8]]

Explanation:

  • np.isnan(a) True where elements are NaN.
  • .any(axis=1) checks if any column in a row contains NaN.
  • ~ flips the mask so that only valid rows remain.

Using np.isreal() + Manual Numeric Check

This method validates each row by checking whether all values are numeric. It works even when the array contains strings or Python objects.

Python 
import numpy as np

a = np.array([[10.5, 22.5, 3.8],
              ["41", "x", "50"]], dtype=object)

m = []
for r in a:
    ok = True
    for v in r:
        s = str(v)
        if s.replace('.', '', 1).replace('-', '', 1).isdigit() == False:
            ok = False
            break
    if ok:
        m.append(r.astype(float))

res = np.array(m)
print(res)

Output
[[10.5 22.5  3.8]]

Explanation:

  • str(v) converts each value to a string so the numeric check never errors.
  • s.replace('.', '', 1).replace('-', '', 1).isdigit() removes one decimal point and one minus sign, then checks if the rest is numeric.
  • Row is valid only if all values pass the numeric check.
  • r.astype(float) converts valid rows to float because every value is confirmed numeric.
Comment

Explore