My entry for https://www.kaggle.com/c/whats-cooking

whats_cooking.py

import json
import argparse

import numpy as np

from sklearn.feature_extraction import DictVectorizer
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import LabelEncoder

parser = argparse.ArgumentParser()
parser.add_argument('--train')
parser.add_argument('--test')
parser.add_argument('--pred')
args = parser.parse_args()

tokenizer = CountVectorizer().build_tokenizer()

def gen_ngrams(x, order = 1):
    for o in range(order):
        for i in range(len(x) - o):
            yield x[i:i+o+1]

def ingredient_bag_of_words(ingredients):
    for ingredient in ingredients:
        tokens = tokenizer(ingredient.lower().replace('-', '_'))
        for ngram in gen_ngrams(tokens, 2):
            yield ' '.join(ngram)

def load(filename):
    D = []
    with open(filename) as f:
        for line in f:
            data = json.loads(line)
            features = {x: 1 for x in ingredient_bag_of_words(data['ingredients'])}
            D.append((data.get('cuisine', None), features, data['id']))
    return D

D_train = load(args.train)
D_test = load(args.test)

featureizer = DictVectorizer()
labelizer = LabelEncoder()

X = featureizer.fit_transform([x[1] for x in D_train])
Y = labelizer.fit_transform([x[0] for x in D_train])

X_test = featureizer.transform([x[1] for x in D_test])

clf = LogisticRegression()
clf.fit(X, Y)
Y_pred = clf.predict(X)
Y_test_pred = clf.predict(X_test)
Y_test_labels = labelizer.inverse_transform(Y_test_pred)

with open(args.pred, 'w') as f:
    for i in range(len(D_test)):
        f.write('{},{}\n'.format(D_test[i][2], Y_test_labels[i]))