Compute Mean Value with MapReduce

This example shows how to compute the mean of a single variable in a data set using mapreduce. It demonstrates a simple use of mapreduce with one key, minimal computation, and an intermediate state (accumulating intermediate sum and count).

Prepare Data

Create a datastore using the airlinesmall.csv data set. This 12-megabyte data set contains 29 columns of flight information for several airline carriers, including arrival and departure times. In this example, select ArrDelay (flight arrival delay) as the variable of interest.

ds = tabularTextDatastore('airlinesmall.csv', 'TreatAsMissing', 'NA');
ds.SelectedVariableNames = 'ArrDelay';

The datastore treats 'NA' values as missing, and replaces the missing values with NaN values by default. Additionally, the SelectedVariableNames property allows you to work with only the selected variable of interest, which you can verify using preview.

preview(ds)
ans=8×1 table
ArrDelay
________

8
8
21
13
4
59
3
11

Run MapReduce

The mapreduce function requires a map function and a reduce function as inputs. The mapper receives blocks of data and outputs intermediate results. The reducer reads the intermediate results and produces a final result.

In this example, the mapper finds the count and sum of the arrival delays in each block of data. The mapper then stores these values as the intermediate values associated with the key "PartialCountSumDelay".

Display the map function file.

function meanArrivalDelayMapper (data, info, intermKVStore)
% Data is an n-by-1 table of the ArrDelay. Remove missing values first:
data(isnan(data.ArrDelay),:) = [];

% Record the partial counts and sums and the reducer will accumulate them.
partCountSum = [length(data.ArrDelay), sum(data.ArrDelay)];
end

The reducer accepts the count and sum for each block stored by the mapper. It sums up the values to obtain the total count and total sum. The overall mean arrival delay is a simple division of the values. mapreduce only calls this reducer once, since the mapper only adds a single unique key. The reducer uses add to add a single key-value pair to the output.

Display the reduce function file.

function meanArrivalDelayReducer(intermKey, intermValIter, outKVStore)
count = 0;
sum = 0;
while hasnext(intermValIter)
countSum = getnext(intermValIter);
count = count + countSum(1);
sum = sum + countSum(2);
end
meanDelay = sum/count;

% The key-value pair added to outKVStore will become the output of mapreduce
end

Use mapreduce to apply the map and reduce functions to the datastore, ds.

meanDelay = mapreduce(ds, @meanArrivalDelayMapper, @meanArrivalDelayReducer);
********************************
*      MAPREDUCE PROGRESS      *
********************************
Map   0% Reduce   0%
Map  16% Reduce   0%
Map  32% Reduce   0%
Map  48% Reduce   0%
Map  65% Reduce   0%
Map  81% Reduce   0%
Map  97% Reduce   0%
Map 100% Reduce   0%
Map 100% Reduce 100%

mapreduce returns a datastore, meanDelay, with files in the current folder.

Read the final result from the output datastore, meanDelay.

ans=1×2 table
Key               Value
____________________    __________

{'MeanArrivalDelay'}    {[7.1201]}

Local Functions

Listed here are the map and reduce functions that mapreduce applies to the data.

function meanArrivalDelayMapper (data, info, intermKVStore)
% Data is an n-by-1 table of the ArrDelay. Remove missing values first:
data(isnan(data.ArrDelay),:) = [];

% Record the partial counts and sums and the reducer will accumulate them.
partCountSum = [length(data.ArrDelay), sum(data.ArrDelay)];
end
%-------------------------------------------------------------------------
function meanArrivalDelayReducer(intermKey, intermValIter, outKVStore)
count = 0;
sum = 0;
while hasnext(intermValIter)
countSum = getnext(intermValIter);
count = count + countSum(1);
sum = sum + countSum(2);
end
meanDelay = sum/count;

% The key-value pair added to outKVStore will become the output of mapreduce