# generate data import numpy as np import random from tqdm import tqdm class Data : # x(t) suite la loi Normale (0, std) OMEGA = [4, 5, 6, 7, 8] TYPE = ["NOR", "UT", "DT", "US", "DS", "CYC"] def __init__(self, mean=30, std=0.05, time=25) : self.mean = mean self.std = std self.time = np.arange(1, time+1) self.v = 0 self.y = [] self.d_min = self.std * 0.1 self.d_max = self.std * 0.3 self.d = np.random.uniform(low=self.d_min, high=self.d_max) self.s_min = self.std * 1.5 self.s_max = self.std * 3 self.s = np.random.uniform(low=self.s_min, high=self.s_max) self.a_min = self.std * 1.5 self.a_max = self.std * 4 self.omega = random.choice(self.OMEGA) self.a = np.random.uniform(low=self.a_min, high=self.a_max) self.choice = [] def typeOf(self, choice) : return max(choice, key=choice.count) def generate(self) : choice = np.random.choice(self.TYPE) for time in tqdm(self.time) : x = np.random.normal(loc=0, scale=self.std) if choice == "NOR" : self.v = 0 y = self.mean + x self.y.append(y) self.choice.append(choice) if choice == "UT" : self.v = 1 y = self.mean + x + self.v * self.d * time self.y.append(y) self.choice.append(choice) if choice == "DT" : self.v = 1 y = self.mean + x - self.v * self.d * time self.y.append(y) self.choice.append(choice) if choice == "US" : self.v = 1 y = self.mean + x + self.v * self.s self.y.append(y) self.choice.append(choice) if choice == "DS" : self.v = 1 y = self.mean + x - self.v * self.s self.y.append(y) self.choice.append(choice) if choice == "CYC" : self.v = 1 y = self.mean + x - self.v * self.a * np.sin(2*np.pi*time/self.omega) self.y.append(y) self.choice.append(choice) return (self.y, choice) class DataController : def __init__(self, sample = 10000) : self.list_y, self.list_choice = [], [] self.sample = sample def generate(self, mean=30, std=0.05, time=25) : self.time = np.arange(1, time + 1) for _ in range(self.sample) : data = Data(mean, std, time) feature, choice = data.generate() self.list_y.append(feature) self.list_choice.append(choice)

import matplotlib.pyplot as plt import seaborn as sns new_data = Data(30, 0.05, 25) feature, type_ = new_data.generate() sns.lineplot(x=new_data.time, y=feature) plt.show() print(f"Type of graph : {type_}")

datacontrol = DataController() datacontrol.generate() list_y, list_choice = datacontrol.list_y, datacontrol.list_choice

sns.lineplot(x=datacontrol.time, y=list_y[30]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[30]}")

list_NOR = [i for i, item in enumerate(list_choice) if item == "NOR"] ind = random.choice(list_NOR) print(ind) sns.lineplot(x=datacontrol.time, y=list_y[ind]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[ind]}")

list_DT = [i for i, item in enumerate(list_choice) if item == "DT"] ind = random.choice(list_DT) print(ind) sns.lineplot(x=datacontrol.time, y=list_y[ind]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[ind]}")

list_UT = [i for i, item in enumerate(list_choice) if item == "UT"] ind = random.choice(list_UT) print(ind) sns.lineplot(x=datacontrol.time, y=list_y[ind]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[ind]}")

list_US = [i for i, item in enumerate(list_choice) if item == "US"] ind = random.choice(list_US) print(ind) sns.lineplot(x=datacontrol.time, y=list_y[ind]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[ind]}")

list_DS = [i for i, item in enumerate(list_choice) if item == "DS"] ind = random.choice(list_DS) print(ind) sns.lineplot(x=datacontrol.time, y=list_y[ind]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[ind]}")

list_CYC = [i for i, item in enumerate(list_choice) if item == "CYC"] ind = random.choice(list_CYC) print(ind) sns.lineplot(x=datacontrol.time, y=list_y[ind]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {list_choice[ind]}")

import pandas as pd from sklearn.model_selection import train_test_split from sklearn.preprocessing import StandardScaler from keras.utils import to_categorical scaler = StandardScaler() def processing(list_y, list_choice) : list_y = np.asarray(list_y) list_choice = np.asarray(list_choice) choice = pd.DataFrame({"choice" : list_choice}) choice = np.asarray(pd.get_dummies(choice)) print(choice) list_y = scaler.fit_transform(list_y) X_train, X_test, y_train, y_test = train_test_split(list_y, choice, random_state=42, test_size=0.2) return X_train, X_test, y_train, y_test

X_train, X_test, y_train, y_test = processing(list_y, list_choice)

x_train = X_train.reshape((X_train.shape[0], X_train.shape[1], 1)) x_test = X_test.reshape((X_test.shape[0], X_test.shape[1], 1))

tmp = x_train[100].reshape((1,25)) sns.lineplot(x=datacontrol.time, y=tmp[0]) plt.ylim((29.5, 30.5)) plt.show() print(f"Type of graph : {y_train[100]}")

tmp

label = ["CYC", "DS", "DT", "NOR", "US", "UT"]

from keras.models import Sequential from keras.layers import Input, Conv1D, MaxPooling1D, Flatten, Activation, Dense # make model class Model : def __init__(self, shape = 25, activation="softmax") : self.model = Sequential() self.model.add(Input(shape=(shape, 1))) self.model.add(Conv1D(filters = 6, kernel_size=2)) self.model.add(MaxPooling1D(pool_size=2)) self.model.add(Conv1D(filters=12, kernel_size=9)) self.model.add(MaxPooling1D(pool_size=2)) self.model.add(Flatten()) self.model.add(Dense(units=6, activation=activation)) # self.model.add(Activation("relu")) self.model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"]) def fit(self, X, y) : self.model.fit(X, y, batch_size = 30, epochs = 100, validation_split = 0.1)

model_softmax = Model() model_softmax.fit(x_train, y_train)

model_sigmoid = Model(activation="sigmoid") model_sigmoid.fit(x_train, y_train)

score = model_softmax.model.evaluate(x_test, y_test, verbose=0) print(f"score softmax : {score[1]}") score = model_sigmoid.model.evaluate(x_test, y_test, verbose=0) print(f"score sigmoid : {score[1]}")