Time series analysis python

time_series
#1

my data plot has no positive or negative trend also there are data of day wise.
Still the mean and std plot is not straight line ? still is it not stationary? confused since visually we cannot say and after dickey fuller plot also getting a bad p value which is .1608

0 Likes

#2

did u handle outliners , also have you tried to apply any transformations to the dataset

0 Likes

#4

Hi ,

Thanks for the reply.
I have used the box plot to remove them also read the articles about time series but still confused, what exactly this transformation do ? And when we train our model on these transformed series how can it predict values based on our series, suggest me some transformation techniques though

0 Likes

#5

hi kunal, transformation techniques are used when the data seems to be highly skewed to achive normality or to reduce variance, when u try dickery fuller test in time series u can observe the p-value if it is more than critical value , u can try to transform data(e.g log, sqrt, sine etc…) and rerun the test for satisfied p-value. Note: after applying transformation u need to revert the obtained result to original scale(re.g for log - 10 power, sqrt- squaring , sine-inverse sine etc.)

0 Likes

#7

Also i have an issues running the predict function on the result from the ARIMA,

Function :
start_index = ‘2015-12-26’
end_index = ‘2016-02-14’
print(results_AR.predict(start=start_index, end=end_index))

ERROR : KeyError: ‘only integers, slices ( : ), ellipsis ( ... ), numpy.newaxis ( None ) and integer or boolean arrays are valid indices’

why is this error and how do i solve ?

0 Likes