We have presented a novel technique of detecting intermittencies in a financial time series of the foreign exchange rate data of U.S.- Euro dollar(US/EUR) using a combination of both statistical and spectral techniques. This has been possible due to Continuous Wavelet Transform (CWT) analysis which has been popularly applied to fluctuating data in various fields science and engineering and is also being tried out in finance and economics. We have been able to qualitatively identify the presence of nonlinearity and chaos in the time series of the foreign exchange rates for US/EURO (United States dollar to Euro Dollar) and US/UK (United States dollar to United Kingdom Pound) currencies. Interestingly we find that for the US-INDIA(United States dollar to Indian Rupee) foreign exchange rates, no such chaotic dynamics is observed. This could be a result of the government control over the foreign exchange rates, instead of the market controlling them.
Wavelet Denoised-ResNet CNN and LightGBM Method to Predict Forex Rate of Change
外汇（Forex）是世界上最大的金融市场。外汇市场的每日交易量远远高于股票和期货市场。因此，建立外汇预测模型对投资者具有重要意义。在本文中，我们提出了一个Wavelet Denoised-ResNet with LightGBM模型来预测五个时间间隔后外汇价格的变化率，以便有足够的时间执行交易。所有的价格都是通过小波变换去噪，并通过计算技术指标形成一个矩阵的30个时间间隔。图像特征是通过将最大值送入ResNet来获得的。最后，将技术指标和图像特征输入LightGBM。我们对5分钟USDJPY的实验表明，该模型优于基线模型，MAE：0.240977x10EXP-3 MSE：0.156x10EXP-6和RMSE：0.395185x10EXP-3。未来25分钟后的准确价格预测为对冲基金算法交易提供了机会窗口。代码可从https://mkhushi.github.io/获得
