Addition of Google Map API , Trees and Train data and End points for the same.

This MR addresses issue #18 (closed) , #16 (closed) ,#15 (closed) .

1. This includes data for trees from extension , db script for it and the endpoint for it.
2. Trains data also has been added to the raw data and the db script for the same has been made.
3. The trains route and trees route has been added to the routes handler.
4. Documentation and test cases has been updated accordingly.

The config.json has been updated with google map api , will pm that in slack.

Just wondering what order of interpolation are we using?

Liner Interpolation , for the dataset we have this seems to work more accurately than other order of interpolations.

but if you have more data points. Linear interpolation doesn't takes those into account. Then how is it more accurate? I think we should use a computation library to get more accurate results.

Everpolate is something that we had in mind , for polynomial interpolation. How ever it does not support extrapolation and the kind of data it takes in the extrapolation example is not supported in our schema. And the linear regression offered in this library has been essentially implemented in our function. http://borischumichev.github.io/everpolate/#poly

Extrapolation, doesn't matter what method you use, is always ambiguous. Now, there is a difference between linear interpolation and linear regression. Linear interpolation finds a point between two points joined by a straight line. Whereas, Linear regression finds a point on a straight line whose slope depends on many points.

See the picture below. The brown line is the linear regression of the data represented by blue circles. Whereas linear interpolation would be joining each consecutive point with a straight line(a polynomial). The second picture shows polynomial interpolation. Clearly, polynomial is more accurate than regression. So, I think it's better to go with polynomial interpolation if the data lies in between. I would also like @bruno-wp to share his views.

Linear Regression:

Interpolation:

@heychirag we just did some calculations using polynomial interpolation and the results are really bad. Have a look at https://runkit.com/embed/rjcp7jlnzk0i. I think we need to have more data points to get the results that are actually useful.

Here is the linear interpolation result for reference. https://runkit.com/embed/9qpyhf7lppy7

So the method in the 2nd figure is not actually polynomial interpolation. Sorry for this. It's actually more closer to either spline or pchip interpolation. The following interpolated results have been calculated using MATLAB for the same data set and test point(2720).

• linear interpolation: 14789.68
• spline interpolation: 14794.34
• pchip interpolation: 14794.22
• linear regression: 14899.14

pchip and spline can be well used for extrapolation too.

extrapolation using pchip:

extrapolation using spline:

The best method depends on the type of data we are dealing with. If we are dealing with data about which we are certain is a non-linear function, then it is better to go with spline or pchip. Here is a good answer that differentiates when to use interpolation and when to use regression.

So if we decide to stick with regression, then I would suggest you to change all the function names because it's quite misleading.

@heychirag we are actually using linear interpolation and not linear regression.

Linear Regression yields some inaccurate values in the lower range even: https://runkit.com/embed/3xjfg2cfrrz5

And the linear regression offered in this library has been essentially implemented in our function.

With reference to above, I was under the impression that we were were using linear regression.

So if we are using linear interpolation, then I would recommend to instead go for spline or pchip interpolation. You can clearly see the difference between spline/pchip and linear interpolation.

I have replaced linear interpolation with spline interpolation.

Did we decide to use flights as generic transport? I might have missed some conversations before.

I didn't get your question , this is just the categories so that if we need we can query and aggregate related stuff.

The idea here is that this piece of data should belong both to the "flights" category and to the "transport" category.

On the one hand, this makes sense, since flights are a form of transport.

On the other hand, I wonder if adding both the category and its super-categories in all pieces of data that belong to that category is an efficient way to organize the sub-categorization information. It seems redundant. It should suffice to describe once that "flights" is a sub-category of "transport".

This is not a new problem in computer science. There is whole field of research concerned about this. Google "Ontology", "Decription Logics", and so on...

Anyway, these concerns are a little bit beyond the scope of our project for now. So, choose whatever feels more natural for you, and let's worry about this later only if it becomes a problem.

This name will pose problems in future.

Yes , I did notice it , thought the name was there in the data set , I should change it i guess.

changed this line in version 2 of the diff

I think we should use delimiter such as a hyphen (-) or underscore (_) over here.

hyphen should be fine they use that on the official website. Will add the same.

changed this line in version 2 of the diff

Hi @Everyone! Just a general comment, as I haven't had time to look into this MR in detail yet. I noticed that this MR addresses several issues (and quite unrelated issues: trains, trees, ...). I strongly recommend that, in the future, you try to submit smaller MRs, and one MR per issue. This has several advantages, including the following:

1. it reduces MR reviewing and acceptance time

2. if you submit a single MR for two issues, a good solution for one of the issues might be blocked, if the MR gets stuck because the reviewers think that it doesn't solve the other issue properly. If 2 separate MRs had been submitted, the one containing the good solution could be accepted quickly.

3. the ability to submit separate and independent MRs is also an indication that your code is modular and capable of scaling independently in different directions. If you only submit big MRs that solve several issues at once, this may be an indication that your code is becoming too monolithic, with interdependencies that prevent independent MRs.

added 1 commit

• 6c52735e - changed names accordingly

Compare with previous version

here too. Check if you have left any other.

changed this line in version 3 of the diff

Changed it

added 1 commit

• d840f120 - changed train names

Compare with previous version

added 1 commit

• ab98ef1b - interpolation changed from linear to spline

Compare with previous version

This MR has conflicts. Could you solve them, so that we can approve it?

Use better names instead of a, b, c and d.

changed this line in version 5 of the diff

Changed parameter names.

Typo: "frieght" --> "freight"

changed this line in version 5 of the diff

Changed it.

Ok. Remember to add tests for trains, flights, trees, and so on...

Added test cases for the same.

approved this merge request

added 1 commit

• 1ad26e63 - added tree test case , changed freight spelling and controller parameter names

Compare with previous version

added 19 commits

• 1ad26e63...fc0287ee - 18 commits from branch aossie:master
• 08c994ce - mr11 merge conflicts resolved

Compare with previous version

@bruno-wp @Kolpa @heychirag The merge conflicts are resolved and the changed required are incorporated.

approved this merge request

approved this merge request

mentioned in commit c05ea726

merged