import pandas as pd
import numpy as np
from sklearn.preprocessing import MinMaxScaler
import matplotlib.pyplot as plt
df = pd.read_csv('/content/processed.csv')
df['Date'] = pd.to_datetime(df['Date'], format="%d-%m-%Y")
print(df['Date'].dtype)
df.index = df['Date']
df.drop('Date', axis=1, inplace=True)
datetime64[ns]

df
PM2.5 CO2 NO2 NH3 CO SO2 O3 AQI
Date
2015-03-04 103.75 42.23 36.98 10.04 1.33 4.65 27.54 289
2015-04-01 106.27 33.67 38.34 11.53 1.08 6.22 27.65 206
2015-04-02 42.82 23.17 35.32 10.17 1.21 6.23 27.60 248
2015-04-03 45.57 0.49 37.37 0.25 0.81 9.23 68.48 111
2015-04-05 85.86 17.20 26.79 12.33 1.14 3.96 27.58 302
... ... ... ... ... ... ... ... ...
2021-06-06 12.81 0.95 53.12 2.45 0.59 6.07 11.68 45
2021-06-07 11.81 0.95 40.84 2.49 0.57 6.04 15.42 34
2021-06-08 14.04 0.95 44.77 2.63 0.57 5.88 11.45 32
2021-06-09 16.26 0.95 49.22 2.01 0.61 6.19 10.09 41
2021-06-10 14.21 0.95 39.15 1.72 0.59 5.59 13.85 33

2224 rows × 8 columns

# Normalize the data
scaler = MinMaxScaler(feature_range=(0, 1))
scaled_data = scaler.fit_transform(df)

# Convert the dataframe to numpy array
data = scaled_data
# Define the percentage of the dataset to be used for training
training_data_len = int(np.ceil(len(data) * 0.8))

# Create the training data
train_data = data[0:training_data_len, :]

# Split the data into x_train and y_train
x_train = []
y_train = []

for i in range(80, len(train_data)):
    x_train.append(train_data[i-80:i, :])
    y_train.append(train_data[i, :])

# Convert x_train and y_train to numpy arrays
x_train, y_train = np.array(x_train), np.array(y_train)

# Create the test data
test_data = data[training_data_len - 80:, :]

# Split the data into x_test and y_test
x_test = []
y_test = data[training_data_len:, :]

for i in range(60, len(test_data)):
    x_test.append(test_data[i-80:i, :])

# Convert x_test to numpy arrays
x_test = np.array(x_test)
from keras.models import Sequential
from keras.layers import Dense, LSTM, Dropout

# Build the LSTM model
model = Sequential()
model.add(LSTM(units=100, return_sequences=True, input_shape=(x_train.shape[1], x_train.shape[2])))
model.add(Dropout(0.2))
model.add(LSTM(units=100, return_sequences=True))
model.add(Dropout(0.2))
model.add(LSTM(units=100))
model.add(Dropout(0.2))
model.add(Dense(units=50))
model.add(Dense(units=8))  # 8 output features

# Compile the model
model.compile(optimizer='adam', loss='mean_squared_error')

# Train the model
history = model.fit(x_train, y_train, epochs=50, batch_size=32, validation_split=0.2)
Epoch 1/50
33/33 [==============================] - 16s 190ms/step - loss: 0.0155 - val_loss: 0.0074
Epoch 2/50
33/33 [==============================] - 6s 173ms/step - loss: 0.0101 - val_loss: 0.0070
Epoch 3/50
33/33 [==============================] - 5s 160ms/step - loss: 0.0096 - val_loss: 0.0067
Epoch 4/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0093 - val_loss: 0.0063
Epoch 5/50
33/33 [==============================] - 6s 191ms/step - loss: 0.0088 - val_loss: 0.0065
Epoch 6/50
33/33 [==============================] - 5s 150ms/step - loss: 0.0089 - val_loss: 0.0061
Epoch 7/50
33/33 [==============================] - 6s 169ms/step - loss: 0.0088 - val_loss: 0.0063
Epoch 8/50
33/33 [==============================] - 5s 159ms/step - loss: 0.0089 - val_loss: 0.0071
Epoch 9/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0085 - val_loss: 0.0060
Epoch 10/50
33/33 [==============================] - 6s 190ms/step - loss: 0.0083 - val_loss: 0.0058
Epoch 11/50
33/33 [==============================] - 6s 179ms/step - loss: 0.0082 - val_loss: 0.0062
Epoch 12/50
33/33 [==============================] - 10s 317ms/step - loss: 0.0084 - val_loss: 0.0070
Epoch 13/50
33/33 [==============================] - 5s 151ms/step - loss: 0.0080 - val_loss: 0.0061
Epoch 14/50
33/33 [==============================] - 8s 242ms/step - loss: 0.0079 - val_loss: 0.0069
Epoch 15/50
33/33 [==============================] - 6s 196ms/step - loss: 0.0079 - val_loss: 0.0063
Epoch 16/50
33/33 [==============================] - 7s 227ms/step - loss: 0.0077 - val_loss: 0.0066
Epoch 17/50
33/33 [==============================] - 5s 162ms/step - loss: 0.0077 - val_loss: 0.0064
Epoch 18/50
33/33 [==============================] - 7s 203ms/step - loss: 0.0076 - val_loss: 0.0060
Epoch 19/50
33/33 [==============================] - 6s 183ms/step - loss: 0.0075 - val_loss: 0.0066
Epoch 20/50
33/33 [==============================] - 6s 188ms/step - loss: 0.0074 - val_loss: 0.0064
Epoch 21/50
33/33 [==============================] - 5s 150ms/step - loss: 0.0075 - val_loss: 0.0065
Epoch 22/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0073 - val_loss: 0.0074
Epoch 23/50
33/33 [==============================] - 6s 196ms/step - loss: 0.0073 - val_loss: 0.0058
Epoch 24/50
33/33 [==============================] - 5s 149ms/step - loss: 0.0073 - val_loss: 0.0058
Epoch 25/50
33/33 [==============================] - 6s 190ms/step - loss: 0.0073 - val_loss: 0.0060
Epoch 26/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0071 - val_loss: 0.0058
Epoch 27/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0069 - val_loss: 0.0062
Epoch 28/50
33/33 [==============================] - 6s 187ms/step - loss: 0.0069 - val_loss: 0.0061
Epoch 29/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0068 - val_loss: 0.0056
Epoch 30/50
33/33 [==============================] - 6s 186ms/step - loss: 0.0067 - val_loss: 0.0054
Epoch 31/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0065 - val_loss: 0.0061
Epoch 32/50
33/33 [==============================] - 5s 162ms/step - loss: 0.0064 - val_loss: 0.0054
Epoch 33/50
33/33 [==============================] - 6s 176ms/step - loss: 0.0065 - val_loss: 0.0061
Epoch 34/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0065 - val_loss: 0.0053
Epoch 35/50
33/33 [==============================] - 6s 191ms/step - loss: 0.0064 - val_loss: 0.0056
Epoch 36/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0063 - val_loss: 0.0051
Epoch 37/50
33/33 [==============================] - 5s 158ms/step - loss: 0.0062 - val_loss: 0.0066
Epoch 38/50
33/33 [==============================] - 6s 178ms/step - loss: 0.0062 - val_loss: 0.0049
Epoch 39/50
33/33 [==============================] - 5s 149ms/step - loss: 0.0060 - val_loss: 0.0050
Epoch 40/50
33/33 [==============================] - 6s 189ms/step - loss: 0.0060 - val_loss: 0.0055
Epoch 41/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0060 - val_loss: 0.0051
Epoch 42/50
33/33 [==============================] - 5s 163ms/step - loss: 0.0061 - val_loss: 0.0044
Epoch 43/50
33/33 [==============================] - 6s 171ms/step - loss: 0.0059 - val_loss: 0.0052
Epoch 44/50
33/33 [==============================] - 5s 151ms/step - loss: 0.0059 - val_loss: 0.0052
Epoch 45/50
33/33 [==============================] - 6s 190ms/step - loss: 0.0059 - val_loss: 0.0052
Epoch 46/50
33/33 [==============================] - 5s 148ms/step - loss: 0.0058 - val_loss: 0.0053
Epoch 47/50
33/33 [==============================] - 6s 170ms/step - loss: 0.0057 - val_loss: 0.0053
Epoch 48/50
33/33 [==============================] - 5s 160ms/step - loss: 0.0058 - val_loss: 0.0058
Epoch 49/50
33/33 [==============================] - 5s 150ms/step - loss: 0.0057 - val_loss: 0.0055
Epoch 50/50
33/33 [==============================] - 6s 193ms/step - loss: 0.0057 - val_loss: 0.0053

# Get the predicted scaled values
predictions = model.predict(x_test)

# Unscale the predicted values
predictions = scaler.inverse_transform(predictions)

# Unscale the true values
y_test = scaler.inverse_transform(y_test)
14/14 [==============================] - 2s 42ms/step

# prompt: accuracy of all 8 columns

from sklearn.metrics import mean_absolute_percentage_error
accuracy_scores = []
for i in range(8):
    accuracy = 100 - mean_absolute_percentage_error(y_test[:, i], predictions[:, i])
    accuracy_scores.append(accuracy)

for i, column in enumerate(df.columns):
    print(f"Accuracy for {column}: {accuracy_scores[i]:.2f}%")
Accuracy for PM2.5: 99.05%
Accuracy for CO2: 93.62%
Accuracy for NO2: 95.24%
Accuracy for NH3: 81.07%
Accuracy for CO: 99.06%
Accuracy for SO2: 99.58%
Accuracy for O3: 99.69%
Accuracy for AQI: 99.54%

# plot for all 8 parameter  as all accuracy above 90 the the graph should overlap actual and accuracy were x axis is date and y axis is value  total 8 plots

# Plot for all 8 parameters
fig, axes = plt.subplots(4, 2, figsize=(15, 15))
axes = axes.flatten()

for i, column in enumerate(df.columns):
    # Plot the actual and predicted values
    axes[i].plot(y_test[:, i], label='Actual', color='blue')
    axes[i].plot(predictions[:, i], label='Predicted', color='orange')

    # Set the title and labels
    axes[i].set_title(f'{column} Prediction')
    axes[i].set_xlabel('Date')
    axes[i].set_ylabel(column)

    # Add legend and grid
    axes[i].legend()
    axes[i].grid(True)

# Adjust the spacing between subplots
plt.tight_layout()

# Show the plot
plt.show()

# Prepare the last 60 days of data for prediction
last_60_days = data[-60:]

# Create an empty list to store the predictions
future_predictions = []

# Predict the next 30 days
for i in range(30):
    # Prepare the input data for prediction
    input_data = last_60_days[-60:]
    input_data = input_data.reshape((1, 60, input_data.shape[1]))

    # Make the prediction
    pred = model.predict(input_data)
    future_predictions.append(pred)

    # Append the prediction to the input data
    last_60_days = np.append(last_60_days, pred, axis=0)

# Convert the predictions to the original scale
future_predictions = np.array(future_predictions)
future_predictions = scaler.inverse_transform(future_predictions.reshape(-1, 8))

# Create a DataFrame for the predictions
future_dates = pd.date_range(start=df.index[-1] + pd.Timedelta(days=1), periods=30)
future_df = pd.DataFrame(future_predictions, index=future_dates, columns=df.columns)

# Display the future predictions
print(future_df)
1/1 [==============================] - 0s 53ms/step
1/1 [==============================] - 0s 47ms/step
1/1 [==============================] - 0s 55ms/step

# prompt: # Display the future predictions
# print(future_df)  in a table

print(future_df.to_string())
                 PM2.5       PM10          NO         NO2          NH3        CO        SO2         O3         AQI
2021-06-11   60.002617  50.883610   19.367136   64.502838   125.114624  0.084118  10.370060  23.871712  111.281906
2021-06-12   63.107674  50.931423   19.667995   67.124435   139.685120  0.134839  10.653792  30.027687  126.617409
2021-06-13   65.961617  50.986816   20.320517   70.815575   164.583359  0.140377  10.899661  34.855602  136.784073
2021-06-14   68.732407  50.799286   21.489349   75.738174   194.399628  0.135283  11.183870  37.846966  143.676620
2021-06-15   71.227882  50.195274   23.231794   82.042854   226.994980  0.156221  11.501744  39.069695  148.632263
2021-06-16   73.960976  49.040543   25.583351   90.032005   262.923340  0.227048  11.918674  39.181587  154.310989
2021-06-17   77.498848  47.231152   28.616997  100.117615   303.442413  0.368728  12.510146  38.920353  163.206650
2021-06-18   82.624298  44.706070   32.421143  112.722176   350.208984  0.594957  13.374964  38.808331  177.737350
2021-06-19   89.576965  41.350803   37.298241  128.602509   404.635071  0.922864  14.569669  39.044239  199.103348
2021-06-20   98.596283  37.118710   43.544445  148.375626   468.300812  1.359627  16.160398  39.463409  228.189209
2021-06-21  109.853798  32.161362   51.309780  172.105316   542.640137  1.891059  18.229544  39.756248  265.564972
2021-06-22  122.630562  26.895773   60.530247  198.991699   627.387085  2.482280  20.787415  39.576397  309.820312
2021-06-23  134.973038  21.945454   70.834549  227.198822   719.213440  3.081508  23.706568  38.524033  356.210358
2021-06-24  145.167496  17.976177   81.305817  253.867737   812.449280  3.627386  26.778667  36.305920  399.092255
2021-06-25  152.073761  15.283709   91.029396  276.791962   900.324036  4.099170  29.735008  32.934376  434.620575
2021-06-26  156.174484  13.837508   99.263039  294.902496   977.809387  4.492974  32.430622  28.939962  463.200653
2021-06-27  158.393494  13.308738  105.696167  308.313965  1042.065552  4.834368  34.739471  24.877388  487.164734
2021-06-28  159.257050  13.299580  110.490067  317.894501  1092.751587  5.136467  36.607742  21.063507  507.349152
2021-06-29  159.326279  13.462250  113.893105  324.476257  1131.428955  5.394811  38.061085  17.656536  523.816406
2021-06-30  158.959076  13.604235  116.233215  328.870331  1160.343018  5.599116  39.177837  14.746899  536.305908
2021-07-01  158.314667  13.566515  117.790405  331.685333  1181.632690  5.753644  40.009975  12.268852  544.747253
2021-07-02  157.514328  13.283374  118.787514  333.384064  1197.182373  5.864694  40.614212  10.147185  549.482605
2021-07-03  156.795959  12.850793  119.387878  334.329285  1208.620361  5.927744  41.081924   8.407581  551.290283
2021-07-04  156.228821  12.299986  119.707817  334.758209  1217.040161  5.951909  41.451111   7.002137  550.914917
2021-07-05  155.806351  11.670095  119.844559  334.858124  1223.279541  5.947795  41.745251   5.861260  548.983765
2021-07-06  155.543106  11.004136  119.859505  334.750854  1227.961426  5.923184  41.986279   4.943182  546.092773
2021-07-07  155.424286  10.335439  119.799706  334.524231  1231.533691  5.885419  42.187820   4.207241  542.689697
2021-07-08  155.445999   9.693343  119.691864  334.229004  1234.319092  5.838931  42.363049   3.623223  539.110535
2021-07-09  155.582382   9.081307  119.555115  333.894836  1236.534912  5.788843  42.515244   3.160630  535.573608
2021-07-10  155.805588   8.510557  119.407600  333.550507  1238.335205  5.738425  42.648857   2.795202  532.205505

from sklearn.metrics import mean_absolute_error, mean_squared_error

# Calculate the evaluation metrics for all 9 columns
for i in range(8):
    mae = mean_absolute_error(y_test[:, i], predictions[:, i])
    mse = mean_squared_error(y_test[:, i], predictions[:, i])
    rmse = np.sqrt(mse)
    print(f"Column: {df.columns[i]}")
    print("Mean Absolute Error (MAE):", mae)
    print("Mean Squared Error (MSE):", mse)
    print("Root Mean Squared Error (RMSE):", rmse)
    print()
Column: PM2.5
Mean Absolute Error (MAE): 20.543749164482495
Mean Squared Error (MSE): 706.9251868186798
Root Mean Squared Error (RMSE): 26.58806474376576

Column: CO2
Mean Absolute Error (MAE): 6.656599771826116
Mean Squared Error (MSE): 122.15371069521923
Root Mean Squared Error (RMSE): 11.052316983113506

Column: NO2
Mean Absolute Error (MAE): 8.842603219178347
Mean Squared Error (MSE): 210.01329653616122
Root Mean Squared Error (RMSE): 14.491835513010807

Column: NH3
Mean Absolute Error (MAE): 42.03412670572062
Mean Squared Error (MSE): 2395.370115467304
Root Mean Squared Error (RMSE): 48.942518483086914

Column: CO
Mean Absolute Error (MAE): 0.41463548268606837
Mean Squared Error (MSE): 0.41688352785812216
Root Mean Squared Error (RMSE): 0.645665182473178

Column: SO2
Mean Absolute Error (MAE): 2.070023189506552
Mean Squared Error (MSE): 8.293230722098933
Root Mean Squared Error (RMSE): 2.8797969932095793

Column: O3
Mean Absolute Error (MAE): 8.264868471187514
Mean Squared Error (MSE): 138.16901845005188
Root Mean Squared Error (RMSE): 11.754531826068272

Column: AQI
Mean Absolute Error (MAE): 29.071826065714294
Mean Squared Error (MSE): 1636.6938253224598
Root Mean Squared Error (RMSE): 40.45607278669619