19
$\begingroup$

I'm delivering an upcoming course on Data Science with the programming/analysis done predominantly in the Wolfram Language. I created the following table to leverage students' familiarity with Lists by way of introducing Associations by way of introducing Datasets. Juxtaposing the various invocations one can't help but be struck with the foresight or "naturalness" of the original language design and subsequent integrations - (IMO an extraterrestrial would be hard-pressed to predict the chronology of sequentially added features, and in particular, to note that all three didn't emerge simultaneously).

My sense is that programming expertise in the Wolfram Language starts with an awareness of its existing (or "likely") functionality followed by inculcating and memorising its basic functional forms. While the documentation is essential for understanding all the details and an overall coherency, for quick reference/comparison/contrast, I think cheatsheets could be more commonly deployed (with no doubt improved visual design - the code follows which could potentially be improved/combined with other users' cheatsheets into a larger, more powerful "Demonstration"?)

At any rate, while an introduction involves becoming familiar with the basic syntax and manipulating structured datasets, the next stage involves applying these to some real-life examples. While there are good texts/datasets available (e.g. Luís Torgo's Data Mining with R: Learning with Case Studies) I'm wondering if anyone has found useful, available datasets from a variety of contexts (health, business, economics, ecology, physics, learning analytics) that might be particularly suitable for showcasing some of the benefits of data science in the Wolfram Language (this might be due to their size, unusual analysis, parallelizability, integration with curated data etc).

SetAttributes[{IOCells, DefCells, InsertIOCells, InsertDefCells}, 
  HoldAll];


DefCells[Set[lhs_, rhs_]] := Row[{
    ExpressionCell[Defer@lhs, "Input", ShowStringCharacters -> True, 
     FontSize -> 16],
    ExpressionCell[" = ", ShowStringCharacters -> False, 
     FontSize -> 16],
    ExpressionCell[lhs = rhs, "Input", ShowStringCharacters -> True, 
     FontSize -> 16]}];

(* Need to split in this way since a Defer wrapper seems to produce \
unexpected formatting - compare:
ExpressionCell[Defer[lhs=<|"a"\[Rule]8,"b"\[Rule]9,"c"\[Rule]10|>],\
"Input"]
 ExpressionCell[Defer[lhs=\[LeftAngleBracket]"a"\[Rule]8,"b"\[Rule]9,\
"c"\[Rule]10\[RightAngleBracket]],"Input"]
 *)

DefCells[R : Set[dataset, Dataset[{_}]]] := (R; Row[{
     ExpressionCell[Defer@dataset, "Input", 
      ShowStringCharacters -> True, FontSize -> 16],
     ExpressionCell[" = ", ShowStringCharacters -> False, 
      FontSize -> 16],
     ExpressionCell["Dataset"[HoldForm[{assoc}]], "Input", 
      ShowStringCharacters -> False, FontSize -> 16]}]);

IOCells[expr_] /; ! 
    FreeQ[Hold@expr, 
     Histogram | ListPlot | ListLinePlot | BarChart3D | BarChart | 
      PieChart | SmoothHistogram] := Grid[{
    {ExpressionCell[Defer@expr, "Input", 
      ShowStringCharacters -> True]},
    {ExpressionCell[expr, "Output", Magnification -> 0.41]}
    }, Frame -> True, Alignment -> Left, 
   Background -> {None, {1 -> GrayLevel[.9], 2 -> White}}];


IOCells[expr_] := Grid[{
    {ExpressionCell[Defer@expr, "Input", 
      ShowStringCharacters -> True]},
    {ExpressionCell[expr, "Output"]}
    }, Frame -> True, Alignment -> Left, 
   Background -> {None, {1 -> GrayLevel[.9], 2 -> White}}];

(* For the same reason above need a specific definition when <|  |> \
notation is in input *)

IOCells[R : KeySelect[assoc_, assoc2_]] := With[
   {t = ToExpression@ToBoxes[assoc2]}, Grid[{
     {ExpressionCell[HoldForm@KeySelect[assoc, t], "Input", 
       ShowStringCharacters -> True]},
     {ExpressionCell[ReleaseHold@R, "Output"]}
     }, Frame -> True, Alignment -> Left, 
    Background -> {None, {1 -> GrayLevel[.9], 2 -> White}}]];

IOCells[R : Dataset[expr_][op_]] := Grid[{
    {ExpressionCell["Dataset"[HoldForm@assoc][op], "Input", 
      ShowStringCharacters -> False]},
    {ExpressionCell[ReleaseHold@R, "Output"]}
    }, Frame -> True, Alignment -> Left, 
   Background -> {None, {1 -> GrayLevel[.9], 2 -> White}}];

(* Since Dataset outputs as grid in the input - means copying \
expression from the grid and evaluating won't work *)


InsertDefCells[defs_List] := 
  ReleaseHold@With[{t = Hold@defs}, Map[DefCells, t, {2}]];
InsertIOCells[exprLs_] := 
  Sequence @@ 
   Map[Row, 
    ReleaseHold@With[{t = Hold@exprLs}, Map[IOCells, t, {4}]], {2}];

ColumnHeadStyle[cheads_List] := 
  ExpressionCell[#, Style[#, {Bold, 18}], 
     ShowStringCharacters -> False] & /@ cheads;


Block[{$PlotTheme = "Minimal"}, Grid[{
   {"List", "Association", "Dataset"} // ColumnHeadStyle,
   {ls = {7, 8, 9}, assoc = <| "a" -> 7, b -> 8, "c" -> 9 |>, 
     dataset = Dataset[{assoc}]} // InsertDefCells,
   {
     {{ls[[All]]}, {Values[assoc], Keys[assoc], Normal[assoc], 
       Lookup[assoc, "b", "No-b"]}, {dataset[1, Values], 
       dataset[1, Keys]}},
     {{(Query@1)[ls], ls[[1]], ls[[2]]}, {assoc[["a"]], assoc["a"], 
       assoc[[Key@b]], assoc@Key@b, 
       Query["a"][assoc]}, {Query[1, Key["a"]][dataset], 
       dataset[1, Key["a"]]}},
     {{ls[[3]], ls[[-1]]}, {assoc[[3]], assoc[[-1]], assoc[a], 
       assoc["b"]}, {dataset[1, 3], dataset[1, -1], 
       dataset[1, Key["d"]]}},
     {{ls[[1 ;; 3]], ls[[1 ;; 3 ;; 2]], Take[ls, {1, 3}], 
       Query[Take[{1, 3}]][ls]}, {assoc[[1 ;; 3]], 
       assoc[[1 ;; 3 ;; 2]], Take[assoc, {1, 3}], 
       Take[assoc, {1, 3, 2}], 
       Query[1 ;; 3][assoc]}, {dataset[1, 1 ;; 3], 
       Query[1, 1 ;; 3][dataset]}},
     {{Query[{1, 3}][ls], ls[[{1, 3}]]}, {Query[{1, 3}][assoc], 
       assoc[[{1, 3}]], assoc[[{"a", "c"}]], 
       KeyTake[assoc, {"a", "c"}]}, {dataset[1, {1, 3}], 
       dataset[1, {"a", "c"}]}},
     {{Cases[ls, _?OddQ], Cases[_?OddQ][ls]}, {Cases[assoc, _?OddQ], 
       Cases[_?OddQ][assoc]}, {dataset[1, Cases[_?OddQ]]}},
     {{Query[Select[OddQ]][assoc], Select[ls, OddQ], 
       SelectFirst[ls, OddQ]}, {Select[assoc, OddQ], 
       Select[OddQ][assoc], 
       SelectFirst[assoc, OddQ]}, {dataset[1, Select[OddQ]], 
       dataset[1, SelectFirst[OddQ]]}},
     {{Position[ls, _?OddQ], 
       FirstPosition[ls, _?OddQ]}, {Position[assoc, _?OddQ], 
       Position[_?OddQ][assoc], 
       FirstPosition[assoc, _?OddQ]}, {dataset[1, Position[_?OddQ]], 
       dataset[1, FirstPosition[_?OddQ]]}},
     {{PositionIndex[ls], 
       Query[PositionIndex][ls]}, {PositionIndex[assoc], 
       Query[PositionIndex][assoc]}, {dataset[1, PositionIndex]}},
     {{Select[Keys@assoc, MemberQ[{"a", "c"}, #] &]}, {KeySelect[
        assoc, MemberQ[{"a", "c"}, #] &], 
       KeySelect[assoc, <| "a" -> True, "c" -> True |>]}, {dataset[1, 
        KeySelect[(# == "a" \[Or] # == "c") &]]}},
     {{f[ls[[3]]], Query[3, f][ls]}, {f[assoc[["c"]]], 
       Query["c", f][assoc]}, {dataset[1, "c", f], 
       dataset[1, 3, f]}},
     {{f /@ ls, Map[f][ls], Query[Map[f]][ls]}, {f /@ assoc, 
       Map[f][assoc]}, {dataset[1, Map@f], dataset[1, All, f]}},
     {{MapIndexed[f, ls], MapIndexed[f][ls]}, {MapIndexed[f, assoc], 
       MapIndexed[f][assoc]}, {dataset[1, MapIndexed[f]]}},
     {{MapAt[f, ls, 2], MapAt[f, 2][ls]}, {MapAt[f, assoc, 2], 
       MapAt[f, 2][assoc], 
       MapAt[f, Key[b]][assoc]}, {dataset[1, MapAt[f, Key@b]]}},
     {{MapAt[f, Normal@assoc, {All, 1}]}, {KeyMap[f, assoc], 
       KeyMap[f][assoc]}, {dataset[1, KeyMap[f]]}},
     {{Reverse /@ Normal@assoc, 
       Association[f /@ Normal@assoc]}, {AssociationMap[Reverse, 
        assoc], AssociationMap[f, assoc]}, {dataset[1, 
        AssociationMap@Reverse], dataset[1, AssociationMap[f]]}},
     {{Sort@ls, Sort[ls, Greater], 
       Query[Sort[#1, Greater] &][ls]}, {Sort@assoc, 
       Sort[assoc, Greater], 
       Query[Sort[#1, Greater] &][assoc]}, {dataset[1, Sort], 
       dataset[1, Sort[#, Greater] &]}},
     {{Total[ls]}, {Total[assoc], Query[Total][assoc], 
       Query[Values /* Total][assoc]}, {dataset[1, Total], 
       dataset[1, All /* Total], dataset[1, Total@*Values]}},
     {{Through[{Min, Max, Total, Variance, Median}@
         ls]}, {Query[{Min, Max, Total, Variance, Median}][
        assoc]}, {dataset[1, {Min, Max, Total, Variance, Median}]}},
     {{Through[{Histogram, BarChart, SmoothHistogram, PieChart, 
          ListLinePlot[#, Filling -> Axis] &}@ls]},
      {Query[{Histogram, BarChart, SmoothHistogram, PieChart, 
          ListLinePlot[#1, Filling -> Axis] &}][
        assoc]}, {dataset[
         1, {Histogram, BarChart, SmoothHistogram, PieChart, 
          ListLinePlot[#1, Filling -> Axis] &}] // Normal}},
     {{Thread[Keys[assoc] -> ls]}, {AssociationThread[
        Keys[assoc] -> ls], 
       AssociationThread[Keys[assoc], 
        ls]}, {dataset[
         1, {AssociationThread[Keys@#, 
            Values@#] &, (Keys@# -> Values@#) & /* 
           AssociationThread}] // Normal}},
     {{Thread[Keys[assoc] -> f /@ Keys[assoc]]}, {AssociationMap[f, 
        Keys[assoc]], 
       Query[AssociationMap[f]@*Keys][assoc]}, {dataset[1, 
        Keys /* AssociationMap[f]]}},
     {{(f[#1, #2, #3] &)[Sequence @@ ls]}, {(f[#a, #1[Key[b]], #c] &)[
        assoc], (f[#1["a"], #1[Key[b]], #1["c"]] &)[assoc]}, {dataset[
        1, All /* (f[#a, #1[Key[b]], #c] &)], 
       dataset[1, (Apply@f)@*Catenate]}},
     {{(f[##1] &)[Sequence @@ ls], (f[##2] &)[
        Sequence @@ ls]}, {((f[##] &)[Sequence @@ Values@#] &)[
        assoc]}, {dataset[1, 
        Values /* (Sequence @@ # &) /* (f[##] &)], 
       dataset[1, (f[##] &)@*(Sequence @@ # &)@*Values]}}
     } // InsertIOCells
   }, Frame -> True, 
  Dividers -> {Thick, {Thick, {None}, Thick}, {3 -> Thick, 8 -> Thick,
      13 -> Thick, 19 -> Thick, 23 -> Thick, 25 -> Thick}}, 
  Alignment -> {Center, Center},
  Background -> {Automatic, {ColorData[17, 4], 
     LightBrown, {LightBlue, LightGreen}}}, 
  ItemStyle -> {Automatic, {19}}, Spacings -> {1, {1, 1, 1, {0.5}}}]]
$\endgroup$
  • $\begingroup$ Not clear what is the question. It's about public datasets? Maybe you can see kaggle.com $\endgroup$ – Murta Jul 28 '14 at 3:05
  • $\begingroup$ In your chart I miss some nice examples on how to change values and append new columns and lines in current List/Associations/DataSet. Unfortunately L-value assignment are not working yet. What make such operations very difficult to Datasets, and make Datasets itself not very interesting to Data Science at the present stage. Another serious limitation is the memory issue, as you can see here. $\endgroup$ – Murta Jul 28 '14 at 3:20
  • $\begingroup$ Yes, datasets for which analysis is made possible or easier using WL capabilities. $\endgroup$ – Ronald Monson Jul 28 '14 at 5:16
  • $\begingroup$ @Murta, I agree, some basic assignments should probably be in the chart - it was already getting a little large although better design would free up some space. Actually another follow-up cheatsheet with assignments/manipulations involving deeper datasets, use of (presumably upcoming) Replace, ReplaceAt etc usage, other aggregation operators and especially GroupBy would be useful. I’m assuming these assignments, and efficient memory management issues will be resolved in upcoming months but IMO of interest will be the ability to refine/test a wider range of algorithms on an “industrial scale”. $\endgroup$ – Ronald Monson Jul 28 '14 at 5:36
  • $\begingroup$ related regarding adding a column to a Dataset mathematica.stackexchange.com/q/51472/66 $\endgroup$ – faysou Nov 17 '14 at 9:06
3
$\begingroup$

I think now this question can be answered with the launch of Wolfram Data Repository. This release blog walks you through the usage. Important fact is you can publish your own data and benefit community and your own goals; the process is free. For example, this very nice dataset and examples ( Atlantic Hurricane Data 1851-2017 ) are published by renowned Seth Chandler with the authorship publication and update dates clearly stated. This is quite neat as it can make a nice addition to your data-scientist resume ;-) Don't forget about very relevant to data science and analogical in nature Wolfram Neural Net Repository (see also release blog). Neural Nets are modern data-operators and are pertinent to the data per se.

Besides free availability on the web with downloadable datasets and usage examples at Wolfram Data Repository site:

enter image description here

the datasets are available from within Wolfram Language as ResourceObject and have access framework:

Search

ResourceSearch["Shakespeare"]

enter image description here

Datasets

meteorites = ResourceData["Meteorite Landings"]

enter image description here

GeoGraphics[{Red,PointSize[0.009],Opacity[0.2],
Point@DeleteMissing[meteorites[All,"Coordinates"]]},
GeoRange->Entity["Country","Australia"],ImageSize->600]

enter image description here

Currently available types are (expanding):

enter image description here

Currently available categories are (expanding):

enter image description here

$\endgroup$
  • $\begingroup$ This is interesting as is the way questions get answered years later as part of some inexorable roll-out. Now the challenge seems to be how to make best use of all these datasets for a given task and/or what combination of tutorials/examples give coverage of all the main methods/idioms. $\endgroup$ – Ronald Monson Sep 20 '18 at 6:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.