Aggregation Estimation for 2D Moving Points

Efficient MaxCount and threshold operators of moving objects

“…We made experimental runs against data sets ranging from 10,000 points to 1,500,000 points. 1 In the following experimental analysis, we measure the percentage error of the estimation algorithm relative to the exact-count algorithm as follows:…”

“…Uniform distributions do not test the ability of the algorithm to adapt. In fact from earlier experiments in [1] we have found that using such a distribution gives great (though meaningless) results. The problem resolves to a system capable (and willing to) model a uniform distribution finding a nearly perfect uniform distribution to model.…”

“…The algorithm differs in that the sweeping algorithm integrates a skew-aware density function over the spatial dimensions of the bucket to obtain the time dependent count function. The density function in the bucket increases accuracy over methods given in Anderson [1], Chen and Revesz [5] while maintaining the same number of buckets. This idea is a crucial improvement because we model the point distribution skew in a bucket, whereas previous methods adapted to skew by increasing the number of buckets or changing their shape and contents.…”

“…Our algorithm uses a sweeping method to evaluate the threshold aggregation operators similar to previous approaches from Chen and Revesz [5], Revesz and Chen [25] and Anderson [1]. The algorithm differs in that the sweeping algorithm integrates a skew-aware density function over the spatial dimensions of the bucket to obtain the time dependent count function.…”

“…In the experiments, buckets divide the space up such that there are 5 to 2 These divisions equate to bucket sizes ranging from 5 to 20 units wide in each dimension. Relative to our previous work [1], this algorithm puts much more space into each bucket creating bigger buckets.…”