erlonidas / predicting_stocks_market

Presentation Scope:

Here we used some standard techniques for time series data analysis. These techniques can be found in the folder named 'Code', and the presentation of results is in the Jupyter file named apresentacao.ipynb.

1) EDA of the dataset:

• Analysis of trends and seasonality in the complete dataset
• Application of Trend and Seasonality tools

2) Fourier at different moments of the IBOV price:

• Application of Fourier in Trend
• Application of Fourier in the Stationary period
• Investigate trend component in the Frequency domain

3) Modeling:

• Application of prediction model in the trend period of raw dataset.
• Application of prediction model in smoothed dataset (SMA 8p)
• Application of prediction model in smoothed dataset (wavelet)
• Compare distance results between prediction and real value across datasets.

The best analysis we had was using the Hot-Winter technique to predict the rest of the downtrend (TF=M30)

In this image, in blue you see all data from IBOV between 10/06/2021 until 01/12/2021 TF = 30 min, used to be in train phase. in red you see the data for test and orange is the prediction used with Hot-Winter technique. There are other few analysis and predictions with others techniques. Checkout in 'advanced analysis' folder.