Calculating Weighted Average with LINQ

问题

My goal is to get a weighted average from one table, based on another tables primary key.

Example Data:

Table1

Key     WEIGHTED_AVERAGE

0200    0

Table2

ForeignKey    Length    Value
0200          105       52
0200          105       60
0200          105       54
0200          105       -1
0200          47        55

I need to get a weighted average based on the length of a segment and I need to ignore values of -1. I know how to do this in SQL, but my goal is to do this in LINQ. It looks something like this in SQL:

SELECT Sum(t2.Value*t2.Length)/Sum(t2.Length) AS WEIGHTED_AVERAGE
FROM Table1 t1, Table2 t2
WHERE t2.Value <> -1
AND t2.ForeignKey = t1.Key;

I am still pretty new to LINQ, and having a hard time figuring out how I would translate this. The result weighted average should come out to roughly 55.3. Thank you.

回答1:

I do this enough that I created an extension method for LINQ.

public static double WeightedAverage<T>(this IEnumerable<T> records, Func<T, double> value, Func<T, double> weight)
{
    double weightedValueSum = records.Sum(x => value(x) * weight(x));
    double weightSum = records.Sum(x => weight(x));

    if (weightSum != 0)
        return weightedValueSum / weightSum;
    else
        throw new DivideByZeroException("Your message here");
}

After you get your subset of data the call looks like this.

double weightedAverage = records.WeightedAverage(x => x.Value, x => x.Length);

This has become extremely handy because I can get a weighted average of any group of data based on another field within the same record.

Update

I now check for dividing by zero and throw a more detailed exception instead of returning 0. Allows user to catch the exception and handle as needed.

回答2:

If you're certain that for each foreign key in Table2 there is a corresponding record in Table1, then you can avoid the join just making a group by.

In that case, the LINQ query is like this:

IEnumerable<int> wheighted_averages =
    from record in Table2
    where record.PCR != -1
    group record by record.ForeignKey into bucket
    select bucket.Sum(record => record.PCR * record.Length) / 
        bucket.Sum(record => record.Length);

UPDATE

This is how you can get the wheighted_average for a specific foreign_key.

IEnumerable<Record> records =
    (from record in Table2
    where record.ForeignKey == foreign_key
    where record.PCR != -1
    select record).ToList();
int wheighted_average = records.Sum(record => record.PCR * record.Length) /
    records.Sum(record => record.Length);

The ToList method called when fetching the records, is to avoid executing the query twice while aggregating the records in the two separate Sum operations.

回答3:

(Answering jsmith's comment to the answer above)

If you don't wish to cycle through some collection, you can try the following:

var filteredList = Table2.Where(x => x.PCR != -1)
 .Join(Table1, x => x.ForeignKey, y => y.Key, (x, y) => new { x.PCR, x.Length });

int weightedAvg = filteredList.Sum(x => x.PCR * x.Length) 
    / filteredList.Sum(x => x.Length);

来源：https://stackoverflow.com/questions/2714639/calculating-weighted-average-with-linq

标签

linq

weighted-average