NKSR Hackathon approaches

You should start by explore.ipynb
what worked best is matching.cpp which finds the nearest neighbour (based on rmse) like knn and imputes it's non nan values into - current row. Since the IV values does not change much in a second, it is supposed to find a temporal order. You need to prepare the data into csv with only call put iv columns without any header for this code to work.
then cubic_spline.ipynb for remaining ipynb (if during matching MAX_ITER is reached or rmse < THRESHOLD)
finally applying savgol filter or pca smoothing made predictions better (bot in final_day/pca_smoothing.ipynb)
final_day/lol.csv also works good, it uses xgb with iteratively improving the completely filled data (initially guessed through matching and cubic spline)
I tried many other things like svi or pchip instead of cubic spline but they did not fit correctly in a many rows.
I also tried using call put parity, training xgb on train data's each column to predict previous column, it works fine but not as good as cubic spline.

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
.vscode		.vscode
final_day		final_day
img		img
.gitignore		.gitignore
ReadMe.md		ReadMe.md
avg.ipynb		avg.ipynb
benchmark.ipynb		benchmark.ipynb
call_put_parity.ipynb		call_put_parity.ipynb
cosh.ipynb		cosh.ipynb
csv_processor.exe		csv_processor.exe
cube_polyfit.ipynb		cube_polyfit.ipynb
cubic_spline.ipynb		cubic_spline.ipynb
explore.ipynb		explore.ipynb
final_day.ipynb		final_day.ipynb
makefile		makefile
makefile_new		makefile_new
matching.cpp		matching.cpp
matching.ipynb		matching.ipynb
matching2.cpp		matching2.cpp
matching_avg.cpp		matching_avg.cpp
mean_median.ipynb		mean_median.ipynb
pchip.ipynb		pchip.ipynb
preprocess.ipynb		preprocess.ipynb
quad_fitting.ipynb		quad_fitting.ipynb
quad_polyfit.ipynb		quad_polyfit.ipynb
svi copy.ipynb		svi copy.ipynb
svi_fitting.ipynb		svi_fitting.ipynb
svi_without_transformation.ipynb		svi_without_transformation.ipynb
test_mean_median.ipynb		test_mean_median.ipynb
training_data.ipynb		training_data.ipynb

Provide feedback