I am new to timeseries analysis and was trying to fit my data into the fb prophet model using the holidays parameter but It was giving a key error on a date which doesn't exist. On renaming the whole holidays column to some other name like super it is fitting but the same code is running on my collegues laptops. Please suggest me a fix to this problem. it is fitting without the holidays parameter.
I've tried reinstalling python on my conda env. I've tried running it on a different IDE. I've updated conda as well
df_holidays_events = pd.read_csv('holidays.csv')
df_holidays = df_holidays_events[['event', 'Date']]
df_holidays.columns = ['holiday', 'ds']
df_holidays.ds = pd.to_datetime(df_holidays.ds)
from fbprophet import Prophet
m = Prophet(changepoint_prior_scale=2.5, holidays=df_holidays)
m.fit(x_train)
This is how x_train looks like
DAY month week year ds y
date
2011-01-30 6 1 4 2011 2011-01-30 481
2011-01-31 0 1 5 2011 2011-01-31 522
This is the df_holidays data frame
holiday ds
0 4th of July 2012-07-04
1 Labor Day 2012-08-31
It should actually fit the model but it isn't fitting it.
