SI_2020/package_location_classifier/classifier.py

from sklearn.tree import DecisionTreeClassifier, plot_tree, export_graphviz, DecisionTreeRegressor
from sklearn.externals.six import StringIO
from IPython.display import Image
import pandas as pd
import numpy as np
import pydotplus
from sklearn.model_selection import train_test_split
from sklearn import metrics

PACKAGE_PLACE_TRESHOLD = {
	"normal": 0.8,
	"freezed": 0.85,
	"fragile": 0.85,
	"flammable": 0.9,
	"keep_dry": 0.8
}


class PackageLocationClassifier():
	def __init__(self):

		data = StringIO()

		cols_names = ["product", "category", "temperature", "humidity", "chance_of_survive", "place_here"]
		feature_cols = ["category", "temperature", "humidity"]

		products = pd.read_csv("package_location_classifier/trainset/trainset.csv", header=0, sep=",", names=cols_names)
		testset = pd.read_csv("package_location_classifier/testset/testset.csv", header=None, sep=",", names=cols_names)
		products = products.round({"chance_of_survive": 1})
		testset = testset.round({"chance_of_survive": 1})
		products.chance_of_survive *= 10
		testset.chance_of_survive *= 10
		test_X = pd.get_dummies(testset[feature_cols])
		test_y = testset.chance_of_survive
		products = products.sample(frac=1)
		X_train = pd.get_dummies(products[feature_cols])
		y_train = products.chance_of_survive
		dummies_names = X_train.columns.tolist()

		# X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.01, random_state=1, shuffle=True)

		clf = DecisionTreeRegressor(ccp_alpha=0.02, min_samples_leaf=5, max_depth=5)
		self.predictor = clf.fit(X_train, y_train)



		y_pred = self.predictor.predict(test_X)

		evaluation = pd.DataFrame({'category': testset.category, 'temperature': testset.temperature , 'humid': testset.humidity ,'Actual': test_y, 'Predicted': y_pred})
		evaluation = evaluation.round({'Actual': 3, 'Predicted': 3})
		evaluation['Prediction_diff'] = abs(evaluation['Actual'] - evaluation['Predicted'])
		print("Prediction differs from actual value by average {}".format(round(evaluation['Prediction_diff'].mean(), 2)))
		# export_graphviz(clf, out_file=data, filled=True, rounded=True, special_characters=True, feature_names=dummies_names)
		# graph = pydotplus.graph_from_dot_data(data.getvalue())
		# graph.write_png('Drzewo.png')
		# Image(graph.create_png())

	def check_if_can_place(self, package, tile):
		category = package.category
		cat_treshold = PACKAGE_PLACE_TRESHOLD[category]
		fields = [[
				  tile.air_temperature,
				  tile.humidity,
				  category == "flammable",
				  category == "fragile",
				  category=="freezed" ,
				  category == "keep_dry",
				  category == "normal"
				  ]]

		quality_of_place = round(self.predictor.predict(fields)[0]/10, 2)
		# print("{} - dopasowanie {}".format(package,quality_of_place))
		# pdb.set_trace()
		if quality_of_place > cat_treshold:
			return True
		return False

if __name__ == '__main__':
	cfer = PackageLocationClassifier()