Predict Brent Oil Price by LSTM

-


#1.importing data from yahoo for 2001 to 2021
df = web.DataReader('BZ=F', data_source='yahoo', start='2001-01-01', end='2021-06-30')
df = df.reset_index()

#2. format date data to appropriate format
df['Date']=pd.to_datetime(df['Date'], format="%b %d, %Y") 
df
DateHighLowOpenCloseVolumeAdj Close
02007-07-3076.52999975.44000275.84999875.7399982575.075.739998
12007-07-3177.16999875.66999875.69999777.0500033513.077.050003
22007-08-0177.05999874.86000177.00000075.3499983930.075.349998
32007-08-0276.20999974.26999775.22000175.7600026180.075.760002
42007-08-0376.00000074.52999975.38999974.7500004387.074.750000
........................
34122021-06-2475.76999774.51999775.37000375.55999826664.075.559998
34132021-06-2576.20999974.94999775.61000176.18000031417.076.180000
34142021-06-2876.58999674.51999776.16000474.68000027711.074.680000
34152021-06-2975.51000273.91000474.59999874.76000227711.074.760002
34162021-06-3075.16999874.77999975.13999974.77999924.074.7799

#3. Plot brent oil price history

#4. Create train and test dataset with 70%-30%
# split into train and test sets
train_size = int(len(scaled_data) * 0.70)
test_size = len(scaled_data) - train_size
train, test = scaled_data[0:train_size, :], scaled_data[train_size:len(scaled_data), :]

#5. build LSTM model (Please search on Google for Long Short Term Model (LSTM))
model = Sequential()
model.add(LSTM(units = 50, return_sequences = True, input_shape = (x_train.shape[1], 1)))
model.add(Dropout(0.2))
model.add(LSTM(units = 50, return_sequences = True))
model.add(Dropout(0.2))
model.add(LSTM(units = 50, return_sequences = True))
model.add(Dropout(0.2))
model.add(LSTM(units = 50))
model.add(Dropout(0.2))
model.add(Dense(units = 1))
model.compile(optimizer = 'adam', loss = 'mean_squared_error')
model.fit(x_train, y_train, batch_size=10,epochs=20)

#6. Validating the loss of train and test
Train Mean Absolute Error: 2.7676871476534477 Train Root Mean Squared Error: 3.5561004954208113 Test Mean Absolute Error: 1.6927817893998272 Test Root Mean Squared Error: 2.3757989199770857


#7. Plot the result

It's very nice and better than expected!!!!

#8. Finally, predict brent oil for 30 days future from 30-Jun-2021. What happened?


Oh my Goodness! the oil price goes down! ^^
Price next (1) days of Oil Brent : 74.98288 Price next (2) days of Oil Brent : 75.1864 Price next (3) days of Oil Brent : 75.18406 Price next (4) days of Oil Brent : 75.039986 Price next (5) days of Oil Brent : 74.830185 Price next (6) days of Oil Brent : 74.60653 Price next (7) days of Oil Brent : 74.392654 Price next (8) days of Oil Brent : 74.193634 Price next (9) days of Oil Brent : 74.00678 Price next (10) days of Oil Brent : 73.82796 Price next (11) days of Oil Brent : 73.653854 Price next (12) days of Oil Brent : 73.48243 Price next (13) days of Oil Brent : 73.31274 Price next (14) days of Oil Brent : 73.1438 Price next (15) days of Oil Brent : 72.97505 Price next (16) days of Oil Brent : 72.80588 Price next (17) days of Oil Brent : 72.63606 Price next (18) days of Oil Brent : 72.46532 Price next (19) days of Oil Brent : 72.293625 Price next (20) days of Oil Brent : 72.12104 Price next (21) days of Oil Brent : 71.94766 Price next (22) days of Oil Brent : 71.77382 Price next (23) days of Oil Brent : 71.59962 Price next (24) days of Oil Brent : 71.42495 Price next (25) days of Oil Brent : 71.25018 Price next (26) days of Oil Brent : 71.07561 Price next (27) days of Oil Brent : 70.90114 Price next (28) days of Oil Brent : 70.726906 Price next (29) days of Oil Brent : 70.55298 Price next (30) days of Oil Brent : 70.37943

Comments

Post a Comment

Many thanks for your comment!

Popular posts from this blog

Determination of source and breakthrough mechanism of water production in a naturally fractured basement reservoir by analyzing water production data

-
Image
Le Minh Vu, Tran Thai Son, Vu Viet Hung Excessive production of water in the naturally fractured basement reservoir has been one of the most intriguing problems in production engineering. Water may appear immediately at the early production stage or breakthrough after months on stream. Available methods to determine the source of produced water are entirely based on chloride content. The stiff diagram is one of the well-known graphical methods to confirm whether it is formation water by integrated various chemical components. With the Stiff diagram, the operators are able to detect formation water in the early days, which apparently assists in quantifying the potential of scale deposit, erosion, and so on. Apart from the water sources, it is crucial to understand the water-out mechanism also via a graphical approach. The characteristic trend of water-oil-ratio and its derivative with time in log-log plot indicates the variety of slopes reflecting different flowing mechanisms and
Read more

Hybrid System of ESP and Gas Lift Application from Conceptual Design, Pilot Test to System Analysis

-
Image
Hybrid System of ESP and Gas Lift Application from Conceptual Design, Pilot Test to System Analysis Artificial lift technology application in heavy oil production has been far-reaching developed in the industry over past decades by persistent efforts to improve the ultimate recovery of this “difficult” hydrocarbon. Heavy oil discovery in a marginal field, Cuu Long Basin, Offshore Vietnam is relatively aberrant and pose challenges to full field development. A series of systematic technical studies have been purposely planned from the first discovery of heavy oil in the wildcat well to the modeling study and facility design to accommodate the viscous fluid over the field life. Apart from the thermal method, pumping technology makes remarkable advance by enlarging the drawdown created over the conventional gas lift in several heavy oil projects. After due consideration, the Electrical Submersible Pump (ESP) was finally decided as the key driver to reinforce well production performanc
Read more

A Comparison of Permeability Prediction Methods Using Core Analysis Data

-
Image
A Comparison of Permeability Prediction Methods Using Core Analysis Data Permeability is one of the most important parameters of reservoir rocks, it illustrates the capacity of a rock to transmit fluids (oil, gas, water) in pore spaces. Permeability data can be obtained from routine core analysis in the laboratory on 1.5 in plugs and sidewall core. However, coring is limited due to the cost issue, so permeability prediction in uncored sections play a significantly important role. The variety of methods developed to estimate permeability using pore-scale such as Kozeny-Carman, Swanson, and Pittman. In fact, those equations are applied individually to estimate permeability. In this research, permeability estimation methods will be used on the same rock (sandstone or carbonate rock) to detect the suitable method for each rock type. Porosity, permeability data from routine core analysis and pore throat size from mercury injection capillary pressure on sandstone and carbonate rocks
Read more