BTCUSDT: PRICE UPDATE
Updated
I would like to share this insight about the BTC price forecast.
I am currently testing a code using 'Jupyter Notebook (python)' and the code generates a 14 day forecast at the 4 hour time frame using LSTM.
However, I twisted the LSTM a little bit by adjusting its hyperparameters based on the model's confidence estimate. The following are the forecasted values and represented by the orange line on the graph:
[[64523.582]
[64701.36 ]
[64890.914]
[65035.395]
[65182.293]
[65359.664]
[65534.016]
[65701.64 ]
[65877.234]
[66049.414]
[66215.58 ]
[66386.92 ]
[66560.5 ]
[66730.13 ]]
Based on the forecast, given the prevailing condition of the market, price will attempt to reach the resistance area. Please note that this forecasting technique using LSTM with 'Hyperparameter Adjustments using Confidence Estimates' is still at the validation period.
I also added Stversteves AR Scatterplot (available in TradingView) to help us see and compare price projections (and I have no intention to disprove any ideas presented here nor determine what could be the best forecasting method). Since AR and LSTM are cousins, and LSTM is more way advance than ARIMA and AR, it would not mean that one is better than the other in terms of generating accurate forecast such as bitcoin price. But of course, neural netwroks has more cool advantages over traditional stats because it could learn from variable associations to generate a more reliable and accurate forecast. However, LSTM still uses randomized values to fine tune its own forecast. And this is the reason why I created the Hyperparameter Adjustment method using confidence estimate to minimize or reduce the random selection of parameters (which I think most of the neural netwroks are doing).
More updates will be posted soon!
Thanks!
I am currently testing a code using 'Jupyter Notebook (python)' and the code generates a 14 day forecast at the 4 hour time frame using LSTM.
However, I twisted the LSTM a little bit by adjusting its hyperparameters based on the model's confidence estimate. The following are the forecasted values and represented by the orange line on the graph:
[[64523.582]
[64701.36 ]
[64890.914]
[65035.395]
[65182.293]
[65359.664]
[65534.016]
[65701.64 ]
[65877.234]
[66049.414]
[66215.58 ]
[66386.92 ]
[66560.5 ]
[66730.13 ]]
Based on the forecast, given the prevailing condition of the market, price will attempt to reach the resistance area. Please note that this forecasting technique using LSTM with 'Hyperparameter Adjustments using Confidence Estimates' is still at the validation period.
I also added Stversteves AR Scatterplot (available in TradingView) to help us see and compare price projections (and I have no intention to disprove any ideas presented here nor determine what could be the best forecasting method). Since AR and LSTM are cousins, and LSTM is more way advance than ARIMA and AR, it would not mean that one is better than the other in terms of generating accurate forecast such as bitcoin price. But of course, neural netwroks has more cool advantages over traditional stats because it could learn from variable associations to generate a more reliable and accurate forecast. However, LSTM still uses randomized values to fine tune its own forecast. And this is the reason why I created the Hyperparameter Adjustment method using confidence estimate to minimize or reduce the random selection of parameters (which I think most of the neural netwroks are doing).
More updates will be posted soon!
Thanks!
Note
Here are the LSTM forecast 4HR starts at 2024-06-2264523.58
64701.36
64890.91
65035.39
65182.29
65359.66
65534.01
65701.64
65877.24
66049.41
66215.58
66386.92
66560.5
66730.13
Note
Here are the refined LSTM forecast at 4hr time frame, 3rd major iteration:64137.79
63970.41
63815.79
63648.15
63492.09
63348.33
63209.99
63069.45
62924.04
62776.04
62629.23
62485.62
62344.49
62203.68
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.