Predicting GPS-based PWV Measurements Using Exponential Smoothing

With the spirit of reproducible research, this repository contains all the codes required to produce the results in the manuscript:

S. Manandhar*, S. Dev*, Y. H. Lee and S. Winkler, Predicting GPS-based PWV Measurements Using Exponential Smoothing, IEEE AP-S Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting, 2019 (* Authors contributed equally).

Please cite the above paper if you intend to use whole/part of the code. This code is only for academic and research purposes.

Code organization

The codes are written in python. The codes are tested in python3 version.

Dependencies

• matplotlib: pip3 install matplotlib
• Tkinter: sudo apt-get install python3-tk
• statsmodels: pip3 install statsmodels
• scipy: pip3 install scipy

Usage

1. python3 pwv_forecasting.py: Run this script to compute the predicted PWV values, based on the historical PWV data. This generated Figure 1 of the paper. You need to change the value of the parameter start_index in the file, to check for different slices of time-series data.
2. python3 grid_search.py: Run this script to obtain the distribution of RMSE values w.r.t. historical data and lead times. This generates Figure 2 of the paper. It saves the figure as ./results/rmse.pdf and the corresponding numpy array as rmse_matrix_for_grid.npy.
3. python3 benchmarking.py: Run this script to obtain the benchmarking results of the different methods. This generates the results in Table 1 of the paper. The results are stored in ./results/comparison.txt file.

Scripts

• benchmarking.py: For a particular past time observation, we compare the performance of various methods for various lead times, and store in a text file.
• grid_search.py: Performs the experiments for various lead times and various historical observations. This is repeated for several times, and the average is reported.
• pwv_forecasting.py: Performs pwv forecasting for sample datapoints and plots the figure with train and test legends
• read_matfile.py: Reads the matlab files, performs pre-processing and returns the data series