Function tf_idf

pattern/vector/__init__.py:703–716

Calculates tf * idf on the vector feature weights (in-place).

(vectors=[], base=2.71828)

cos = cosine_similarity

def tf_idf(vectors=[], base=2.71828): # Euler's number
    """ Calculates tf * idf on the vector feature weights (in-place).
    """
    df = {}
    for v in vectors:
        for f in v:
            if v[f] != 0:
                df[f] = df[f] + 1 if f in df else 1.0
    n = len(vectors)
    s = dict.__setitem__
    for v in vectors: 
        for f in v: # Modified in-place.
            s(v, f, v[f] * (log(n / df[f], base)))
    return vectors

tfidf = tf_idf