The data is daily data, that means a single row represents the turtle which is found in a specific day and specific capture site.
Our task is to predict the number of turtles in 2019 weekly not daily until week number 44. first group the 1998-2018 data by week from week 1-52 for each year .
What I did was first create a seprate data for each capture site and add week column. then calculate the week number starting from 2002.
week_number = (year-2002) * 42 + week
Then group the data by week_number count for instance if the data has 3 same week_number for capturesite_0 then the value for that week_number will be 3. Finally by adding zero for the weeks with no turtles and considering the data as a time series data for each capturesite train an ARIMA model and predict for the next 44 weeks.
yeah it's normal. the training data is recorded from 1998-2018, and you have to predict for the number of turtles in 2019 grouped by week.
Please can you explain more , grouped by week as how??
The data is daily data, that means a single row represents the turtle which is found in a specific day and specific capture site.
Our task is to predict the number of turtles in 2019 weekly not daily until week number 44. first group the 1998-2018 data by week from week 1-52 for each year .
What I did was first create a seprate data for each capture site and add week column. then calculate the week number starting from 2002.
week_number = (year-2002) * 42 + week
Then group the data by week_number count for instance if the data has 3 same week_number for capturesite_0 then the value for that week_number will be 3. Finally by adding zero for the weeks with no turtles and considering the data as a time series data for each capturesite train an ARIMA model and predict for the next 44 weeks.
Yes. It is normal.