##Description The recipeSuggestionAI is a simple AI, feeded and trained with a dataset over 13.000 recipes. There is for every recipe a list of ingredients and the belonging preparation. So, the AI response with a recipe, they can found in the dataset. The response including the belonging preparation and the list of ingredients. The
##Requirements
- nltk
- numpy
- pickle
- pandas
- keras
##Getting started Just declare a new instance of mealDishAi and execute the run()-function:
mealAI = mealDishAi()
mealAI.run()##neural-network-architecture The neural network is build with the open-source framework keras. I used a multilayer perceptron with 3 layers:
- first layer: 256 neurons
- second layer: 128 neurons
- third layer: output.
As optimizer, adam is used and for the activation-functions, I use relu. This works fine for simple results, but there is a lot of scope for improvements.