Recommendation algorithms

NearestNeighbors.kt

fun List<Number>.getNearestNeighbors(posibleNeighbors: List<List<Number>>, numberOfNeighbors: Int): List<List<Double>> {
    return posibleNeighbors.mapIndexed { index, posibleNeighbor -> index to this.pearsonCorrelationWith(posibleNeighbor) }
            .sortedBy { it.second }
            .take(numberOfNeighbors)
            .map { posibleNeighbors[it.first].map { it.toDouble() } }

}

PearsonCorrelation.kt

fun Array<Number>.sum(): Double {
    return this.map { it.toDouble() }.reduce { acc, new -> acc + new }
}

fun Array<Number>.sumSq(): Double {
    return this.map { it.toDouble() * it.toDouble() }.reduce { acc, new -> acc + new }
}

fun Array<Number>.pearsonCorrelationWith(otherArray: Array<Number>): Double {
    assert(this.size == otherArray.size)
    val sumX = this.sum()
    val sumY = otherArray.sum()

    val sumXSq = this.sumSq()
    val sumYSq = otherArray.sumSq()

    val pSum = this.zip(otherArray) { x, y -> x.toDouble() * y.toDouble() }.reduce { acc, new -> acc + new }

    val num = pSum - (sumX * sumY / this.size)
    val den = Math.pow((sumXSq - Math.pow(sumX, 2.0) / this.size) * (sumYSq - Math.pow(sumY, 2.0)), 0.5)

    return if (den == 0.0) 1.0 else num/den
}