Prediction system and prediction method used for Internet short and medium term medication consultation business volume

Abstract

The invention discloses a prediction system and a prediction method used for an Internet short and medium term medication consultation business volume. The system comprises a data acquisition unit, aprediction unit and an uploading unit, wherein the data acquisition unit is used for collecting historical business volume data stored in a server and an online doctor number; the prediction unit is used for receiving data provided by the data acquisition unit, calling a prediction model to carry out model training and acquiring prediction result data; and the uploading unit is used for uploadingthe prediction result data to the server. The prediction method comprises the following steps of (1) using an LSTM prediction method to predict on non-holidays, and predicting non-holiday data after holidays through removing holiday data; and (2) using an improved exponential smoothing method to predict on holidays. Doctors in the next day or even a few days can be rapidly, accurately and reasonably scheduled, and a doctor manpower allocation advice is provided.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and specifically relates to a forecasting system and a forecasting method for medium and short-term medication consulting business volume on the Internet. Background technique [0002] The micro consultation terminal is a telemedicine service terminal set up in a pharmacy store. The micro consultation terminal is composed of a touch-type all-in-one high-definition camera, a camera, and a call device integrated with an IP phone, and provides access to national licensed pharmacist qualifications. A team of professional pharmacists and doctors with certificates and years of experience in pharmacies and hospitals provide remote pharmaceutical medical services. It can ensure that licensed pharmacists and doctors have online real-time consultations during the business hours of pharmacies, provide remote prescription review services and electronic prescription circulation services for or...

AI Technical Summary

Problems solved by technology

[0005] There are many commonly used prediction methods, including empirical method, time series prediction method, exponential smoothing method, regression analysis method, artificial neural network prediction method, support vector prediction method, etc. Among them, artificial neural network needs to use a large number of samples to train the model and the training speed is relatively fast. Slow; the ARIMA model requires the sequence to be a stationary sequence; the empirical method requires high labor costs; the exponential smoothing method and regression analysis method are relatively simple, but the accuracy is not high; the time series forecasting method requires complete historical data and is not suitable for seasonal changes business volume forecast

In general, a single forecasting method has its limitations. Under the condition that the amount of data is not huge, due to the randomness and seasonality of the business volume, as well as changes in the number of pharmacy equipment due to climate, holidays, and the number of online doctors and so on, the prediction stability is not enough

[0006] Based on the above problems, in the invention patent CN 107862555A forecasting system and method based on exponential smoothing, an improved exponential smoothing method is proposed to predict the sales data of a certain period of time in the future, but the accuracy of the light exponential smoothing forecast is not high

[0007] In the short-term traffic flow prediction based on the integration of LSTM and gray model in the paper, a traffic flow prediction method integrating LSTM model and GM model is improved, and the two prediction results are weighted by the method of dynamic weight, but this method There are two problems: one is that the forecasted business volume on holidays is higher than the actual business volume, and the other is that the forecasted volume after holidays is generally lower than the actual business volume. No matter how weighted, the predicted value is still lower than the actual value

Prediction accuracy is still not high

Images