CODEWITHSUNDEEP
rmachine-learningtime-seriesdata-analysisdata-manipulation
R.A
R.A

Reputation: 111

How to analyse time series which contains multiple sources (e.g. weather data time series of multiple weather stations)

I am trying to build a machine learning model to predict VAR1 for each neighborhood across time. Using time series data (Year and month). However, the data contains many neighborhoods (which are the basis of the analysis). So each neighborhood will be repeated 3 yrs * 12 months= 36 times.

I need to merge this data with other datasets. All the other datasets have the same number of areas and have Year but do not have month.

I need help on how to go about joining these datasets together and analysis. I am working in R.

On joining the datasets, I will try to transfer the rows in the other datasets into columns so I would have less instances for each area.

Here's an example of some the head of some of the datasets (in R): 

head(df)
  Year     Month          District         Neighborhood          Gender       VAR1             
1 2017    January             1                 1                   M          2000
2 2017    January             1                 2                   M          350
3 2017    January             1                 3                   M          700
4 2017    January             1                 4                   M          400
5 2017    January             2                 5                   M          1000
6 2017    January             2                 6                   M          200


tail(df)
      Year  Month          District           Neighborhood      Gender       VAR1                
10577 2015 December            10                   69              F         200
10578 2015 December            10                   70              F         1000
10579 2015 December            10                   71              F         500
10580 2015 December            10                   72              F         350
10581 2015 December            10                   73              F         300
10582 2015 December            99                   99              F         770

I need help on two things:

First, I need to know how I can merge the example above with the other datasets which don't have month.

I am stuck on how to go about doing EDA and analysing this dataset and would appreciate help here.

Upvotes: 1

Views: 357

Answers (2)

GeekyDad
GeekyDad

Reputation: 75

Several models (ARIMA, HOLT etc.,) then using Ensemble model.

Upvotes: 0

moon
moon

Reputation: 571

You can build separate models for each source then simply concatenate them together. Here is a quick video for this.

Upvotes: 2

Related Questions