Added notebook

code/wilshire_5000/nn.py

@@ -6,8 +6,10 @@ import wilshire
 import tensorflow_addons as tfa
 from statsmodels.tsa.arima.model import ARIMA
 
+
+### Fonctions d'activations ###
 def snake(x):
-    return(x+(tf.math.sin(20*x)**2)/20)
+    return(x+(tf.math.sin(30*x)**2)/30)
 def sinus(x):
     return(tf.math.sin(x))
 def sinus_cosinus(x):
@@ -16,11 +18,14 @@ def swish(x):
     return(x*tf.math.sigmoid(x))
 
 
-#activations = [tf.keras.activations.relu,swish,sinus_cosinus,sinus,snake]
-activations = [snake]
 
 
 def prepare_data(filename="WILL5000INDFC.csv"):
+    """
+    Prepare data by preprocessing, normalizing and cutting it in train and test sets
+    Return x and y train and test sets, as well as the maximum for later plots and the index separating both sets
+
+    """
     df_train,df_test,index = wilshire.preprocess(filename)
     x_train = np.arange(df_train.shape[0])
     maximum = np.max(x_train)
@@ -35,6 +40,10 @@ def prepare_data(filename="WILL5000INDFC.csv"):
     return x_train,x_test,y_train,y_test,maximum,index
 
 def arima_pred(y_train,y_test,orders=[[2,1,1],[2,2,1],[3,1,1],[2,1,2]],n=5):
+    """
+    Computes the ARIMA errors (mse) for several orders to compare with the article
+
+    """
     mse=[]
     for order in orders :
         
@@ -55,6 +64,10 @@ def arima_pred(y_train,y_test,orders=[[2,1,1],[2,2,1],[3,1,1],[2,1,2]],n=5):
 
 
 def create_model(activation):
+    """
+    Create the neural network with the requested activation function
+
+    """
     model =  tf.keras.Sequential()
 
     model.add(tf.keras.layers.Dense(1,input_shape=[1,],activation=activation))
@@ -68,8 +81,12 @@ def create_model(activation):
     model.summary()
     return model
 
-def training_testing(n=5,activations = [tf.keras.activations.relu,swish,sinus_cosinus,sinus,snake]):
-    x_train,x_test,y_train,y_test,maximum,index = prepare_data(filename="WILL5000INDFC2.csv")
+def training_testing(n=5,activations = [tf.keras.activations.relu,swish,sinus_cosinus,sinus,snake],epochs = 50):
+    """
+    Trains models and computes means and std of test errors on n tries for each activation function requested.
+
+    """
+    x_train,x_test,y_train,y_test,maximum,index = prepare_data(filename="WILL5000INDFC.csv")
     models = []
     errors_train,errors_test = [],[]
     mean_y_train,mean_y_test,std_y_test=[],[],[]
@@ -81,7 +98,7 @@ def training_testing(n=5,activations = [tf.keras.activations.relu,swish,sinus_co
         for k in range(n):
 
             model = create_model(activation)
-            model.fit(x_train,y_train, batch_size=1, epochs=50)
+            model.fit(x_train,y_train, batch_size=1, epochs=epochs)
 
             y_pred_test = model.predict(x_test)
             y_pred_train = model.predict(x_train)
@@ -101,25 +118,28 @@ def training_testing(n=5,activations = [tf.keras.activations.relu,swish,sinus_co
     return models,errors_train,errors_test
 
 
-def final_plot(models,errors_test,arima_err,activations=["ReLU","Swish","Sinus Cosinus","Sinus","Snake"]):
-    x_train,x_test,y_train,y_test,maximum,index = prepare_data(filename="WILL5000INDFC2.csv")
-    x = np.arange(9000)
+def final_plot(models,errors_test,arima_err,activations=["ReLU","Swish","Sinus Cosinus","Sinus","Snake"],orders_ARIMA = ["[2,1,1]","[2,2,1]","[3,1,1]","[2,1,2]"]):
+    """
+    Prints the results to compare with the table of the article and plot the same plot as the article
+    """
+    x_train,x_test,y_train,y_test,maximum,index = prepare_data(filename="WILL5000INDFC.csv")
+    x = np.arange(9000) ## 9000 data points bring us to ~2031 to try and predict future data
     x_n = x / maximum
-    future_preds = models[-1].predict(x_n)  ## Calculated with a website the number of working days between 01-06-2020 and 01-01-2024
+    future_preds = models[-1].predict(x_n) 
 
-    #x=np.arange(df_train.shape[0]+df_test.shape[0]+908)
     y_true = np.concatenate((y_train,y_test))
     x_cut = np.arange(x_train.shape[0]+x_test.shape[0])
 
     print("----- ARIMA Test MSE -----")
-    orders_ARIMA = ["[2,1,1]","[2,2,1]","[3,1,1]","[2,1,2]"]
-    # for k in range(len(orders_ARIMA)):
-    #     print("ARIMA"+orders_ARIMA[k]+" : "+str(arima_err[k]))
+    for k in range(len(orders_ARIMA)):
+        print("ARIMA"+orders_ARIMA[k]+" : "+str(arima_err[k]))
     
     print("----- DNN Test MSE -----")
-    
     for k in range(len(activations)):
         print("DNN "+activations[k]+" : "+str(errors_test[k]))
+
+
+    ### PLOT ###
     plt.figure()
     plt.plot(x_cut,y_true,label="True data")
     plt.plot(x,future_preds,label="Predictions")
@@ -131,22 +151,16 @@ def final_plot(models,errors_test,arima_err,activations=["ReLU","Swish","Sinus C
     plt.show()
 
 
-x_train,x_test,y_train,y_test,maximum,index = prepare_data()
-#mse = arima_pred(y_train,y_test)
-# mse=[]
-# # models,errors_train,errors_test = training_testing(n=1,activations=[snake])
-# # models[0].save("Snake20a")
-# models=[]
-# errors_test=[]
-# models.append(tf.keras.models.load_model("Snake30a"))
-# print(mse,errors_test)
-# final_plot(models,errors_test,mse,activations=[])
+
 
 def plot_all_a(a=["1","10","20","30","100"]):
+    """
+    Plots the varying a values plot by loading pre-existing models (they are uploaded on GitHub)
+    """
     models=[]
     for param in a :
         models.append(tf.keras.models.load_model("Snake"+param+"a"))
-    x_train,x_test,y_train,y_test,maximum,index = prepare_data(filename="WILL5000INDFC2.csv")
+    x_train,x_test,y_train,y_test,maximum,index = prepare_data(filename="WILL5000INDFC.csv")
     x = np.arange(9000)
     x_n = x / maximum
     y_true = np.concatenate((y_train,y_test))

code/wilshire_5000/wilshire.py

@@ -6,9 +6,8 @@ import matplotlib.pyplot as plt
 
 def parser(path):
     df = pd.read_csv(path,na_values='.')
-    #df = df.interpolate()
+    #df = df.interpolate()  ### Interpolate or dropna for bank holidays
     df = df.dropna().reset_index(drop=True)
-    #df = df.drop(labels=np.arange(1825)) ### To obtain the same graph than in the article
     return(df)
 
 def preprocess(path):
@@ -17,13 +16,9 @@ def preprocess(path):
     df_normalized = df[:]
     df_normalized["WILL5000INDFC"]=df_normalized["WILL5000INDFC"]/np.max(df_normalized["WILL5000INDFC"])
     index_train = int(df_normalized[df_normalized["DATE"]=="2020-01-31"].index.array[0])
-    # df.plot()
-    # plt.show()
+
     df_train = df_normalized[:index_train]
-    df_test = df_normalized[index_train+1:index_train+85]
+    df_test = df_normalized[index_train+1:index_train+85] #Between 02-01 and 05-31
 
-    # df_train.plot()
-    # df_test.plot()
-    # plt.show()
     return(df_train,df_test,index_train)