Task description Solution
Guessing the cuisine of recipes according to their
ingredients
Project ”pattern recognition” - Pre-presentation
Michael Jobst
October 26, 2015
1/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Task description
▶ guessing the cuisine a recipe belongs to from the list of its
ingredients
▶ example data: {’ingredients’: [’romaine lettuce’, ’black olives’,
’grape tomatoes’, ’garlic’, ’pepper’, ’purple onion’,
’seasoning’, ’garbanzo beans’, ’feta cheese crumbles’],
’cuisine’: ’greek’, ’id’: 10259}
2/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Task description
▶ guessing the cuisine a recipe belongs to from the list of its
ingredients
▶ example data: {’ingredients’: [’romaine lettuce’, ’black olives’,
’grape tomatoes’, ’garlic’, ’pepper’, ’purple onion’,
’seasoning’, ’garbanzo beans’, ’feta cheese crumbles’],
’cuisine’: ’greek’, ’id’: 10259}
▶ task size:
▶ 39774 recipes in the whole dataset
▶ 20 cuisines (→ classes)
▶ 6081 unique ingredients (→ features)
▶ on average 10.8 ingredients per recipe
2/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Approach for solving
1. convert data (simple import from JSON)
2. feature extraction
3/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Approach for solving
1. convert data (simple import from JSON)
2. feature extraction
2.1 building sparse feature vectors
2.2 feature selection (χ2
-test. . . )
3/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Approach for solving
1. convert data (simple import from JSON)
2. feature extraction
2.1 building sparse feature vectors
2.2 feature selection (χ2
-test. . . )
3. selecting and tuning a good classifier
3/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Approach for solving
1. convert data (simple import from JSON)
2. feature extraction
2.1 building sparse feature vectors
2.2 feature selection (χ2
-test. . . )
3. selecting and tuning a good classifier
▶ SVM
▶ Random tree forests
3/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Approach for solving
1. convert data (simple import from JSON)
2. feature extraction
2.1 building sparse feature vectors
2.2 feature selection (χ2
-test. . . )
3. selecting and tuning a good classifier
▶ SVM
▶ Random tree forests
4. evaluation
3/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients
Task description Solution
Approach for solving
1. convert data (simple import from JSON)
2. feature extraction
2.1 building sparse feature vectors
2.2 feature selection (χ2
-test. . . )
3. selecting and tuning a good classifier
▶ SVM
▶ Random tree forests
4. evaluation
Questions to evaluate
▶ Which classifier yields the best results and how to train it
efficiently?
▶ Is there some interesting statistical information?
3/3 Michael Jobst Guessing the cuisine of recipes according to their ingredients