US 11,989,663 B2
Prediction method, prediction apparatus, and computer-readable recording medium
Takuya Takagi, Kawasaki (JP); Hiroaki Iwashita, Tama (JP); Keisuke Goto, Kawasaki (JP); and Kotaro Ohori, Chuo (JP)
Assigned to FUJITSU LIMITED, Kawasaki (JP)
Filed by FUJITSU LIMITED, Kawasaki (JP)
Filed on Jun. 17, 2020, as Appl. No. 16/904,305.
Claims priority of application No. 2019-126209 (JP), filed on Jul. 5, 2019.
Prior Publication US 2021/0004698 A1, Jan. 7, 2021
Int. Cl. G06N 5/04 (2023.01); G06N 5/01 (2023.01); G06N 20/00 (2019.01)
CPC G06N 5/04 (2013.01) [G06N 5/01 (2023.01); G06N 20/00 (2019.01)] 9 Claims
OG exemplary drawing
 
1. A non-transitory computer-readable recording medium storing therein a prediction program that causes a computer to execute a process comprising:
storing in a storage unit a plurality of pieces of rule information that is generated based on training data and is indicated by an association of a combination of an attribute value, from among attribute values, of an attribute, from among a plurality of attributes, with a label, from among labels, so that the plurality of pieces of rule information is usable to obtain a predicted value of a target label for an attribute value, from among the attribute values;
receiving input data including an attribute value of a part of an attribute, from among the plurality of attributes, and a predetermined value of the target label, from among the labels;
generating a tree structure data based on the plurality of pieces of rule information and a predetermined order condition for the plurality of attributes, the tree structure data being obtained by aggregating the plurality of pieces of rule information, the tree structure data including an attribute value, from among the attribute values, as a branch,
the predetermined order condition indicates a process order in a manufacturing process and an attribute value of an attribute indicates an action including an observed value or a controlled setting value in the manufacturing process;
acquiring a likelihood of obtaining the predetermined value of the target label, based on the predetermined order condition for a relationship among the plurality of attributes in the tree structure data, when an attribute value of a predetermined attribute, from among the plurality of attributes, that is other than the part of the attribute in the input data, is set to a certain attribute value;
determining based on the likelihood, when predicting a value of a label for the certain attribute value using the plurality of pieces of rule information, a degree of contribution of the certain attribute value to make the value of the label for the certain attribute value become the predetermined value of the target label;
determining the certain attribute value of the predetermined attribute, based on the degree of contribution, and storing the certain attribute value of the predetermined attribute in the storage unit; and
outputting the certain attribute value of the predetermined attribute for the target label.