import pandas as pd
import seaborn as sb
import matplotlib.pyplot as plt
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import f1_score

CLASS = "Aktivitätskategorie"
FEATURES = [
    "Distanz km", 
    "Ø Herzfrequenz",
    "Maximale Herzfrequenz",
    "Aerober TE", 
    "Ø Geschwindigkeit km/h",
    "Maximale Geschwindigkeit km/h",
    "Kalorien",
    "Zeit"
]

#Define class for all the classifiers and set the instance variables for the values
class Classifier:
    def __init__(self, model):
        self.model = model
        self.f1s = [[[] for j in range(len(FEATURES)) ] for i in range(len(FEATURES))]
        self.f1_means = [[] for i  in range(len(FEATURES))]
        self.f1_medians = [[] for i  in range(len(FEATURES))]

#Function to load the dataframe
def load_dataframe():
    file = "data/Sport_Daten.csv"
    try:
        columns = FEATURES.copy() # Copy the Values and dont use the same object as FEATURES
        columns.append(CLASS)
        data_frame = pd.read_csv(file, usecols = columns, sep=";") # Sep is required because from excel the csv is exported with ;
        return data_frame
    except:
        return None
  
#Function to create a lineplot
def show_lineplot(name_of_classifier, data_frame):
    sb.lineplot(data=data_frame, x="X", y="Y", hue="Features")
    plt.title(name_of_classifier + ":\n Performance mit Anzahl Features")
    plt.xlabel("Anzahl Features")
    plt.ylabel("F1-Mittelwert")
    plt.savefig("plots/lineplot_" + "_".join(name_of_classifier.split()) + ".png")
    plt.show()
 
#Main function    
def main():
    #Load Data Frame and check if it was successful
    data_frame = load_dataframe()
    if data_frame is None:
        return
    else:
        print("✅ Load Dataframe")
        
    
    #Create new Column and transform Activity column to numbers
    labelencoder = LabelEncoder()
    data_frame[CLASS] = labelencoder.fit_transform(data_frame[CLASS])
    
    #Create Random Forest Classifier
    random_forest = RandomForestClassifier(random_state=0)
    # Create Decision Tree Classifier
    decision_tree = DecisionTreeClassifier(random_state=0)
    #Create K-Neighbors Classifier
    k_neighbors = KNeighborsClassifier(n_neighbors=5)
        
    #Define Set with Classifiers and name
    classifiers = {
        "Random Forest Classifier": Classifier(model = random_forest),
        "Decision Tree Classifier": Classifier(model = decision_tree),
        "K-Nearest-Neighbors Classifier": Classifier(model = k_neighbors),
    }
    
    #Itterate trough all combinations and calculate binary Value of integer with Array of Features
    for i in range(1,2**len(FEATURES)):                 #from 1 to 2^number of Features
        binary_value = bin(i)[2:].zfill(len(FEATURES))  #Calculate binary value and fill with zeros
        features = []
        for j in range(len(FEATURES)):      #Itterate trough all positions in binary code
            if int(binary_value[j]) == 1:   #Check if binary Value is 1 / If feature should be used
               features.append(FEATURES[j]) 
        
        #Define dataframes
        y = data_frame[CLASS]
        X = data_frame[features]
        y = pd.get_dummies(y)
    
        #Split data in test and training data
        X_train , X_test , y_train , y_test = train_test_split(X, y,test_size=0.2, random_state=0)
    
        #For all classifiers train the model, predict values and calculate f1-score
        for index, (name, classifier) in enumerate(classifiers.items()):
            classifier.model.fit(X_train.values , y_train.values)
            pred = classifier.model.predict(X_test.values)
            f1 = f1_score(y_test.values , pred, average='macro')
            
            #For each Features which was used to train write f1-score in matrix
            for j in range(len(FEATURES)):
                if int(binary_value[j]) == 1:
                    classifier.f1s[j][len(features)-1].append(f1)
            
        #print progress on console, overwrite the current value
        progress = i * 100 / (2**len(FEATURES)-1)
        print(f"\r Progress: {progress:.2f}%", end="", flush=True)
    
    
    #Make calculations for all classifiers
    for name, classifier in classifiers.items():
        #Calculate Average of F1 Score
        for i in range(len(FEATURES)):
            for j in range(len(FEATURES)): 
                f1s_for_nr_features = []                
                f1s_for_nr_features.extend(classifier.f1s[i][j]) 
                mean = sum(f1s_for_nr_features) / len(f1s_for_nr_features)
                classifier.f1_means[i].append(mean)
                   
        #Create set for all the values in lineplot 
        rows = []
        for index_feature, feature in enumerate(FEATURES):
            for x in range(len(classifier.f1_means[index_feature])):
                rows.append({
                    "Features": feature,
                    "X": x+1,
                    "Y": classifier.f1_means[index_feature][x]
                })
             
        #Create Dataframe and hand it to the lineplot function  
        line_plot_dataframe = pd.DataFrame(rows)
        show_lineplot(name, line_plot_dataframe)
        
#Start script            
if __name__ == "__main__":
    main()