I am working on runoff prediction. I completed missing data and removed outliers. When it comes to normalization, applying a single normalization method (such as MinMaxScaler, StandardScaler, or Power Transform) does not yield satisfactory accuracy (e.g., r2_score = [0.5, 0.6]). However, when I apply multiple normalization methods sequentially (for example, first logarithm, then MinMax, then Power Transform), the results improve significantly (approximately r2_score = 0.95). I experimented with multiple stations and different models like CatBoost, SVM, MLP, LSTM, etc., and the results were consistent in this regard.

I want to pre-process and merge 10-K filings from multiple companies over a span of ten years for data visualization. The files I have downloaded are in .txt format. How can I download them in a format that can be processed?