Hi SubaAravind--
I used the 2010 and 2020 decennial Census datasets for cities and townships in Michigan, which represent the comprehensive enumeration of the population conducted by the U.S. Census Bureau every 10 years.
I ended up using a tool that applied Getis Ord Gi * for my analysis, which is available in ArcPro (not ArcGIS online). This stat identifies where high or low values cluster in space, so it was a good fit for use in my project because my dataset consisted of both positive and negative values representing the percent change in population for areas between 2010 and 2020. I've included that calculation below, using the city of Detroit as an example:
((Detroit population in 2020 - Detroit population in 2010) / Detroit population in 2010)
If you want to study the distribution of population, you may need to standardize (or "normalize") your data if the polygons within your area of interest differ in terms of (areal) size. This standardization will allow you to make accurate comparisons of the distribution of your population that are meaningful and consistent.
You could always standardize your data by calculating the population density of a given polygon (representing a city, county, etc). To do this, you would divide the population total for that polygon by its total area (in square km). This would give you the number of persons per square kilometer for a dataset of polygons. I believe the hotspot tool you are using in AGOL also provides an option to normalize using the "Divide by" parameter, but it consumes credits, so I always just execute this step in ArcPro myself as a workaround to avoid consuming credits.
From there--depending on the best cluster statistic to apply for your specific project--you could use AGOL's hotspot tool or choose to use another clustering statistic that's available in ArcPro/ArcMap (assuming you have access to ArcPro or ArcMap). If you're interested in learning about all the types of clustering algorithms that ESRI has available for spatial analyses, you can read about them here.
Note: many of these tools require you select the input parameters that are essentially used to model the spatial relationships in your data. This article provides a thorough discussion of how to conceptualize these spatial relationships and choose the appropriate parameters for your project. This is important because the parameters you choose will influence the output of the analysis. I had to play around with different options before finding the one that helped me achieve the goal I mentioned in my post above.
One tool that does NOT require you to select these input parameters is the "Optimized Hotspot Analysis" tool in ArcPro. Whereas the traditional hotspot analysis gives you control over the spatial parameters, the optimized version tries to make some intelligent choices for some of the parameters for you with algorithms that organically interrogate the structure of the data.
Hope that helps, good luck!