Research Article

Load Forecasting Method Based on Improved Deep Learning in Cloud Computing Environment

Algorithm 1

Pseudocode of data partition algorithm.
Parameter implication:
 Input: is the data set; is the maximum number of sample data objects in the grid cell.
 Output: Partitions is the data partition obtained after the data set is divided.
 Begin
(1)Obtain multidimensional data space through sample data set .
(2)The KD-tree algorithm is used to divide the multidimensional data space into multiple grid cells with relatively balanced size and no coincidence.
(3)The sample data objects are allocated to grid cells, and then the number of sample data objects contained in each grid cell is calculated.
(4)Initialize an empty Queue, add data space to the Queue, and initialize an empty grid cell set .
(5)Pop up the Queue header element from the queue, and calculate the number of sample data objects contained in .
(6)If, then
 Add to ;
(7)If, then
 Calculate the variance of each dimension of the data object in the -dimensional space in the spatial area , select the dimension with the largest variance as the segmentation dimension, divide into two subspace areas and with an equal number of data objects, and then add and to the queue to wait for further division.
(8)If Queue is empty, then
 The result of spatial meshing of data set is set ;
Otherwise Skip to step 5.
(9)According to the divided spatial grid set , all data partitions of the data set can be obtained.
 End