Method and system for estimating number of orders

Abstract

The invention discloses a method and system for estimating the number of orders, and the method comprises the steps: constructing a prediction model through the combination of a multi-layer feedforward (BP) neural network and a gradient boosting decision tree (GBDT) neural network, and obtaining direct influence factor parameters of orders and indirect influence factor parameters of the orders; performing word segmentation and sentiment analysis on the indirect influence factor parameters of the orders to obtain sentiment values of the indirect influence factor parameters of the orders; inputting the emotion values of the direct influence factor parameters of the orders and the indirect influence factor parameters of the orders into the constructed prediction model, carrying out pre-estimation processing through a BP neural network and a GBDT neural network in the prediction model, and outputting a first order pre-estimation quantity value and a second order pre-estimation quantity value respectively, and through merging processing of the prediction model, obtaining an order estimation data value. Therefore, according to the embodiment of the invention, the order quantity is accurately estimated, and the estimation accuracy is improved.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and system for estimating order quantity. Background technique [0002] With the development of computer network technology, various applications can be realized through the computer network. For example, a trading platform for goods or services is set up on the computer network side, and the user terminal is connected to the trading platform through the computer network, and the goods or services are booked through the trading platform. In order to better serve the user terminal, the trading platform of goods or services needs to estimate some information, such as estimating the reservation order of the goods or services, so as to prepare the goods or services in advance. [0003] At present, when the trading platform estimates the number of orders, there are two methods, which are explained below. [0004] The first method is the subjective prediction method. The ...

AI Technical Summary

Problems solved by technology

[0006] However, there are also obvious disadvantages in adopting these two methods

For the first method, when estimating the number of orders, there is a certain degree of subjectivity, the historical multi-dimensional information related to the trading platform related to the order is not fully mined, and the external factors related to the volatility of the order are not considered in place, so that There is a large difference between the estimated order quantity and the actual order quantity, and the estimation accuracy is low, which cannot adapt to the opportunities and challenges brought by the order fluctuation of the trading platform

For the second method, although the neural network model is used to estimate the order, the accuracy of the estimate still cannot meet the requirements

For example, the linear prediction model based on time series mainly focuses on the relationship between time and the information involved in the order, while ignoring other influences. When the linear prediction model encounters the actual problem of nonlinear relationship, the accuracy of the prediction will be reduced. decline

For example, if the prediction model of BP neural network is used alone, the model is slow in the convergence process, and it is likely to fall into local minimum and poor stability, resulting in low prediction accuracy.

Images