Spaces:
Sleeping
Sleeping
Upload series_time.ipynb
Browse files- series_time.ipynb +154 -0
series_time.ipynb
ADDED
@@ -0,0 +1,154 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "code",
|
5 |
+
"execution_count": null,
|
6 |
+
"metadata": {},
|
7 |
+
"outputs": [],
|
8 |
+
"source": [
|
9 |
+
"import streamlit as st\n",
|
10 |
+
"import pandas as pd\n",
|
11 |
+
"import numpy as np\n",
|
12 |
+
"import torch\n",
|
13 |
+
"import torch.nn as nn\n",
|
14 |
+
"import matplotlib.pyplot as plt\n",
|
15 |
+
"from sklearn.preprocessing import MinMaxScaler\n",
|
16 |
+
"\n",
|
17 |
+
"# Cargar los datos de los dos CSV\n",
|
18 |
+
"file1 = 'PARCIAL-AGUA-_2_.csv'\n",
|
19 |
+
"file2 = 'PARCIAL-AGUA-_3_.csv'\n",
|
20 |
+
"\n",
|
21 |
+
"data1 = pd.read_csv(file1)\n",
|
22 |
+
"data2 = pd.read_csv(file2)\n",
|
23 |
+
"\n",
|
24 |
+
"# Convertir la columna 'FECHA' a objetos datetime y filtrar por años\n",
|
25 |
+
"data1['FECHA'] = pd.to_datetime(data1['FECHA'])\n",
|
26 |
+
"data2['FECHA'] = pd.to_datetime(data2['FECHA'])\n",
|
27 |
+
"\n",
|
28 |
+
"filtered_data1 = data1[data1['FECHA'].dt.year >= 2007]\n",
|
29 |
+
"filtered_data2 = data2[data2['FECHA'].dt.year >= 2007]\n",
|
30 |
+
"\n",
|
31 |
+
"# Combinar los valores de ambos conjuntos de datos\n",
|
32 |
+
"combined_values = np.concatenate([filtered_data1['VALOR-LS-CF-N'].values, filtered_data2['VALOR-LS-CF-N'].values]).reshape(-1, 1)\n",
|
33 |
+
"\n",
|
34 |
+
"# Seleccionar la variable objetivo y escalar los valores\n",
|
35 |
+
"scaler = MinMaxScaler()\n",
|
36 |
+
"scaled_values = scaler.fit_transform(combined_values)\n",
|
37 |
+
"\n",
|
38 |
+
"# Dividir los datos escalados en los conjuntos de datos originales\n",
|
39 |
+
"scaled_values1 = scaled_values[:len(filtered_data1)]\n",
|
40 |
+
"scaled_values2 = scaled_values[len(filtered_data1):]\n",
|
41 |
+
"\n",
|
42 |
+
"# Función para crear ventanas deslizantes\n",
|
43 |
+
"def sliding_windows(data, seq_length):\n",
|
44 |
+
" x, y = [], []\n",
|
45 |
+
" for i in range(len(data) - seq_length):\n",
|
46 |
+
" x.append(data[i:i + seq_length])\n",
|
47 |
+
" y.append(data[i + seq_length])\n",
|
48 |
+
" return np.array(x), np.array(y)\n",
|
49 |
+
"\n",
|
50 |
+
"# Preparar las ventanas deslizantes para cada conjunto de datos\n",
|
51 |
+
"seq_length = 4\n",
|
52 |
+
"x_train, y_train = sliding_windows(scaled_values1, seq_length)\n",
|
53 |
+
"x_test, y_test = sliding_windows(scaled_values2, seq_length)\n",
|
54 |
+
"\n",
|
55 |
+
"# Convertir a tensores de PyTorch\n",
|
56 |
+
"trainX = torch.Tensor(x_train)\n",
|
57 |
+
"trainY = torch.Tensor(y_train)\n",
|
58 |
+
"testX = torch.Tensor(x_test)\n",
|
59 |
+
"testY = torch.Tensor(y_test)\n",
|
60 |
+
"\n",
|
61 |
+
"# Definir el modelo LSTM\n",
|
62 |
+
"class LSTM(nn.Module):\n",
|
63 |
+
" def __init__(self, input_size, hidden_size, num_layers, output_size):\n",
|
64 |
+
" super(LSTM, self).__init__()\n",
|
65 |
+
" self.hidden_size = hidden_size\n",
|
66 |
+
" self.num_layers = num_layers\n",
|
67 |
+
" self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True)\n",
|
68 |
+
" self.fc = nn.Linear(hidden_size, output_size)\n",
|
69 |
+
"\n",
|
70 |
+
" def forward(self, x):\n",
|
71 |
+
" h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size)\n",
|
72 |
+
" c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size)\n",
|
73 |
+
" out, _ = self.lstm(x, (h0, c0))\n",
|
74 |
+
" out = self.fc(out[:, -1, :])\n",
|
75 |
+
" return out\n",
|
76 |
+
"\n",
|
77 |
+
"# Aquí puedes definir otros modelos si lo deseas\n",
|
78 |
+
"# ...\n",
|
79 |
+
"\n",
|
80 |
+
"# Interfaz de Streamlit\n",
|
81 |
+
"st.title('Predicción de Series de Tiempo')\n",
|
82 |
+
"st.sidebar.title('Parámetros del Modelo')\n",
|
83 |
+
"\n",
|
84 |
+
"model_type = st.sidebar.selectbox('Selecciona el modelo', ('LSTM', 'Otro Modelo'))\n",
|
85 |
+
"num_epochs = st.sidebar.slider('Número de épocas', 100, 500, 200)\n",
|
86 |
+
"learning_rate = st.sidebar.number_input('Tasa de aprendizaje', 0.001, 0.1, 0.01, 0.001)\n",
|
87 |
+
"\n",
|
88 |
+
"if model_type == 'LSTM':\n",
|
89 |
+
" input_size = 1\n",
|
90 |
+
" hidden_size = 50\n",
|
91 |
+
" num_layers = 2\n",
|
92 |
+
" output_size = 1\n",
|
93 |
+
"\n",
|
94 |
+
" model = LSTM(input_size, hidden_size, num_layers, output_size)\n",
|
95 |
+
"\n",
|
96 |
+
" criterion = nn.MSELoss()\n",
|
97 |
+
" optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
|
98 |
+
"\n",
|
99 |
+
" if st.sidebar.button('Entrenar y Predecir'):\n",
|
100 |
+
" for epoch in range(num_epochs):\n",
|
101 |
+
" model.train()\n",
|
102 |
+
" outputs = model(trainX)\n",
|
103 |
+
" optimizer.zero_grad()\n",
|
104 |
+
" loss = criterion(outputs, trainY)\n",
|
105 |
+
" loss.backward()\n",
|
106 |
+
" optimizer.step()\n",
|
107 |
+
" if (epoch+1) % 100 == 0:\n",
|
108 |
+
" st.write(f'Epoch [{epoch+1}/{num_epochs}], Loss: {loss.item():.4f}')\n",
|
109 |
+
"\n",
|
110 |
+
" model.eval()\n",
|
111 |
+
" train_predict = model(trainX)\n",
|
112 |
+
" test_predict = model(testX)\n",
|
113 |
+
"\n",
|
114 |
+
" train_predict = scaler.inverse_transform(train_predict.detach().numpy().reshape(-1, 1))\n",
|
115 |
+
" trainY_plot = scaler.inverse_transform(trainY.numpy().reshape(-1, 1))\n",
|
116 |
+
" test_predict = scaler.inverse_transform(test_predict.detach().numpy().reshape(-1, 1))\n",
|
117 |
+
" testY_plot = scaler.inverse_transform(testY.numpy().reshape(-1, 1))\n",
|
118 |
+
"\n",
|
119 |
+
" fig, ax = plt.subplots(figsize=(12, 6))\n",
|
120 |
+
" ax.plot(filtered_data1['FECHA'].values[seq_length:seq_length+len(trainY)], trainY_plot, label='Datos de entrenamiento')\n",
|
121 |
+
" ax.plot(filtered_data1['FECHA'].values[seq_length:seq_length+len(trainY)], train_predict, label='Predicciones de entrenamiento')\n",
|
122 |
+
" ax.plot(filtered_data2['FECHA'].values[seq_length:seq_length+len(testY)], testY_plot, label='Datos de prueba')\n",
|
123 |
+
" ax.plot(filtered_data2['FECHA'].values[seq_length:seq_length+len(testY)], test_predict, label='Predicciones de prueba')\n",
|
124 |
+
" ax.set_xlabel('Fecha')\n",
|
125 |
+
" ax.set_ylabel('VALOR-LS-CF-N')\n",
|
126 |
+
" ax.set_title('Predicciones con LSTM')\n",
|
127 |
+
" ax.legend()\n",
|
128 |
+
" ax.grid(True)\n",
|
129 |
+
" st.pyplot(fig)"
|
130 |
+
]
|
131 |
+
}
|
132 |
+
],
|
133 |
+
"metadata": {
|
134 |
+
"kernelspec": {
|
135 |
+
"display_name": "Python 3",
|
136 |
+
"language": "python",
|
137 |
+
"name": "python3"
|
138 |
+
},
|
139 |
+
"language_info": {
|
140 |
+
"codemirror_mode": {
|
141 |
+
"name": "ipython",
|
142 |
+
"version": 3
|
143 |
+
},
|
144 |
+
"file_extension": ".py",
|
145 |
+
"mimetype": "text/x-python",
|
146 |
+
"name": "python",
|
147 |
+
"nbconvert_exporter": "python",
|
148 |
+
"pygments_lexer": "ipython3",
|
149 |
+
"version": "3.12.4"
|
150 |
+
}
|
151 |
+
},
|
152 |
+
"nbformat": 4,
|
153 |
+
"nbformat_minor": 2
|
154 |
+
}
|