Best way to use Time Series Forecasting

Hey guys! I have two questions please (pictures and outline below)

BACKGROUND:

Completed the Time Series Forecasting: https://au.mathworks.com/help/deeplearning/ug/time-series-forecasting-using-deep-learning.html

Using my own timeseries data.

HOPE via USE OF MATLAB EXAMPLE

I found and used the Chickpox example as it does what I require = Capture time series data and make a prediction in the future.

QUESTIONS given END GOAL

End Goal: Make a prediction for time period X for Data Points A, B, and C

Help required:

1. Getting a prediction outside the known data set, and
2. Managing more than one data point per 'month'

QUESTION / PROBLEM 1

Actual: Lesson helps to predict and train for data within the known data set

Goal: Predict after the known data set

Visually.... the tutorial delivers the following:

But what I'd like is this:

WONDERING...

I did play arund with this although not 100% sure I've grabbed the right data or trainin set.

>> How can I manipulate the code from within Chickpox example for training on the whole dataset, to predict values for X month(s) into the future and be sure I'm getting predictions. Am I using the right logic / network here?

For example: "From the total dataset, the predictions for the next six months are.... *insert Matlab Magic*"

QUESTION / PROBLEM 2

Say this data set wasn't just a number of total cases + months, but each month with counts for cases by Total Adult + Male under 18 + Female under 18 (3 readings per Month / Time indicator)

I've been able to get through this (in a way) with 2 data points, although its getting hinky with 3+. Ideally, I'll have several. 3 kinda works, but it's a mess.

Predictions goes into an array and can 'lose' the order of the respective value is came from.

I.e. the data set has

• month XX
• adults #'s
• boys cases #
• girl cases #

Array for results can be adult prediction boy prediction gir prediction but later one gets funky to being girl prediction boy prediction adult prediction

IS THIS:

• Due to uneven number of readings in data set?
• Do I just need to be mindful of the order of operation when declaring variables through the journey?
• Something else... (see B)

B) Just stop what I'm doing because something else is better when attempting time series forecasting with mulitple values per time point (that all relate to each other)

[Timothy Rijsdijk]

That is my exact question. How do you use the data for future predictions?

Did you find the answer?