CN117557289A - Restaurant data modeling method and system based on intelligent marketing scene - Google Patents
Restaurant data modeling method and system based on intelligent marketing scene Download PDFInfo
- Publication number
- CN117557289A CN117557289A CN202311547883.3A CN202311547883A CN117557289A CN 117557289 A CN117557289 A CN 117557289A CN 202311547883 A CN202311547883 A CN 202311547883A CN 117557289 A CN117557289 A CN 117557289A
- Authority
- CN
- China
- Prior art keywords
- data information
- data
- neural network
- catering
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000003062 neural network model Methods 0.000 claims abstract description 115
- 238000013528 artificial neural network Methods 0.000 claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 45
- 238000012545 processing Methods 0.000 claims abstract description 43
- 238000010606 normalization Methods 0.000 claims abstract description 12
- 230000008859 change Effects 0.000 claims abstract description 11
- 238000007405 data analysis Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 21
- 235000013361 beverage Nutrition 0.000 claims description 9
- 235000013305 food Nutrition 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 7
- 230000003213 activating effect Effects 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 238000010200 validation analysis Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 230000002779 inactivation Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000000306 recurrent effect Effects 0.000 claims description 3
- 230000006872 improvement Effects 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 208000025174 PANDAS Diseases 0.000 description 1
- 208000021155 Paediatric autoimmune neuropsychiatric disorders associated with streptococcal infection Diseases 0.000 description 1
- 240000000220 Panda oleosa Species 0.000 description 1
- 235000016496 Panda oleosa Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/12—Hotels or restaurants
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Biomedical Technology (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Human Resources & Organizations (AREA)
- Probability & Statistics with Applications (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a restaurant data modeling method and a restaurant data modeling system based on an intelligent marketing scene, wherein the method firstly divides restaurant data information into three types, and performs normalization processing to obtain characteristic information data of first type data information and second type data information; combining the acquired characteristic information data with third-class data information to form a catering data information sample set; training a catering data information sample set through an LSTM neural network to obtain a neural network model; finally, optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, continuously optimizing the neural network model, and predicting the catering data information of a plurality of days in the future through the optimized neural network model; the invention provides a method for updating and optimizing the neural network model in the catering industry, which avoids the situation that the predicted value of the neural network model is more and more inaccurate due to the change of the mode in the catering industry and the change of the dining habits of clients.
Description
Technical Field
The invention relates to a restaurant industry data processing technology, in particular to a restaurant data modeling method and system based on an intelligent marketing scene.
Background
Along with the annual increase of the people consumption level, the scale of the catering industry is also continuously enlarged; in a strong catering market competition, catering enterprises face a double challenge: the cost is continuously increased, and the profit is gradually reduced; modeling of the catering data is performed, sales or profits of catering enterprises are predicted, and the catering enterprises can plan supply chains, inventory management and human resources better, so that service cost is optimized, production loss is reduced, the enterprises can formulate more targeted marketing strategies, more clients are attracted, and service growth is promoted.
However, the data modeling task of the catering industry faces many challenges, mainly including complex data to be processed, complex feature engineering, insufficient data quality, update problems and adaptability problems of the established model, and the like; therefore, it is necessary to find a new method for modeling data in the catering industry, and optimize and update the new method to solve the above-mentioned problems.
Disclosure of Invention
The invention aims to overcome the problems existing in the prior art and greatly improve the technical effect on the basis of the prior art; firstly, the invention provides a restaurant data modeling method based on an intelligent marketing scene, which comprises the following steps:
firstly) acquiring restaurant data information in nearly three years; the dining data information comprises: first class data information, second class data information and third class data information; the first type of data information is merchant dish sales data information, and comprises: the total quantity information of dishes and the corresponding quantity of each dish variety; the second type of data information is client dining data information, and comprises the following steps: old customers and new customers have dinner times; the third type of data information is profit information corresponding to the first type of data information and the second type of data information in a period of time;
secondly), respectively carrying out normalization processing on the first type data information and the second type data information by taking each day as a unit to acquire characteristic information data of the first type data information and the second type data information; combining the characteristic information data of the normalized first type data information and the normalized second type data information with the third type data information to form a restaurant data information sample set { (X) i ,Y i ) Wherein X is i Representing a vector composed of characteristic information data of the first type of data information and the second type of data information, Y i Representation and vector X i Corresponding profit information;
third), the obtained restaurant data information sample set { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain trained neural network model, wherein X i As an input sequence to the LSTM neural network, Y i A target sequence corresponding to the LSTM neural network; the food and beverage data information sample set to be obtained { (X) i ,Y i ) The training process of training by inputting the LSTM neural network is as follows: a) Sample set { (X) i ,Y i ) Inputting into LSTM neural network to obtain model output; b) Comparing the output of the model with the target sequence, and calculating a loss function value; c) Calculating gradients using a back propagation method and updating model parameters with an optimizer to reduce the value of the loss function; d) Repeating the steps until reaching the specified training condition, stopping training and obtaining a neural network model;
fourthly, predicting the catering data information of the first day in the future through a trained neural network model, and acquiring the catering data information of the actual day according to the management condition of the actual day; then, comparing the predicted dining data information with the dining data information of the actual day, if the predicted dining data information is different from the dining data information of the actual day; optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, and continuously optimizing the neural network model;
fifthly), predicting dining data information of a plurality of days in the future through an optimized neural network model; the prediction method comprises the following steps: firstly, predicting restaurant data information of a first day in the future through an optimized neural network model; then, predicting the catering data information of the second future day by taking the predicted catering data information of the first future day as the data of the training set; obtaining dining data information of multiple days in the future through repeated prediction; and after the catering data information of a plurality of days in the future is acquired, recovering the neural network model before prediction.
In particular, the acquiring restaurant data information within the last three years comprises: according to the actual operating conditions of catering companies, acquiring catering data information in the last three years, classifying the catering data information, and dividing the catering data information into: first class data information, second class data information and third class data information.
Specifically, the normalizing the first class data information and the second class data information respectively includes: converting the first type data information and the second type data information into characteristic information data which is easy to compare by adopting a characteristic scaling normalization processing mode; the characteristic data information of the first type data information and the second type data information together form a data set { X } i (wherein X is i A vector formed by characteristic information data of the first type data information and the second type data information; regarding the data set composed of the third type of data information as { Y } i },Y i Corresponding profit information; the catering data information sample set consisting of the first type of data information, the second type of data information and the third type of data information is { (X) i ,Y i )}。
Specifically, the to-be-obtained restaurant data information sample set { (X) i ,Y i ) Training the input LSTM neural network includes: the LSTM neural network is a variant of a Recurrent Neural Network (RNN) with the ability to memorize long and short-term information; the LSTM neural network is used for processing the catering data information, so that the characteristic information of the catering data information at different times can be learned, and the characteristic information can be used for predicting future catering data information; when restaurant data information is processed through LSTM neural network, LSTM modelThe method comprises the following steps: a) The time step of the LSTM model is set to 7, the input dimension is set to 4, and the output dimension is set to 1; b) The activating function of the LSTM layer is a ReLU function, and the activating function of the output layer is a sigmoid function; c) Setting the first, second and third LSTM hidden layer neurons as 256, 256 and 128 respectively; meanwhile, the hidden layer neurons are deleted randomly by adopting Dropout to prevent overfitting, and the random inactivation rate is set to be 0.3; d) Training an LSTM neural network model through tensorfilow, wherein a loss function of the model uses MSE, the learning rate of the model is set to be 0.01, and the initial value of the bias parameter is set to be 0; the batch size, epoch of the model were set to 100 and 200, respectively.
Specifically, the specified conditions until training is reached include: the specified conditions are conditions for the LSTM neural network to reach a stop training, including: achieving a fixed epoch value, no further reduction in validation loss, no further improvement in validation performance, performance meeting target and time constraints.
Specifically, the if the predicted dining data information and the dining data information of the actual day are different includes: the predicted dining data information is the prediction of the trained neural network model on dining data information of the first day in the future, and the dining data information of the actual day refers to the acquisition of the dining data information of the actual day consistent with the predicted date according to the actual business situation; if the predicted dining data information and the dining data information of the actual day are different, the following descriptions are provided: instability of restaurant data information, insufficient generalization capability of a model and unknown future information; meanwhile, the neural network model before prediction is optimized through the catering data information of the actual day different from the predicted catering data information, so that the stability of the neural network model for prediction of the catering industry can be increased, and the optimization of the neural network model is realized.
Specifically, the recovering the neural network model before prediction after acquiring the catering data information of a plurality of days in the future includes: training the neural network by taking the predicted value as a sample set, and predicting future catering data information; the predicted value is not catering data information of the actual day, but is a data embodiment of the neural network model; training the neural network by taking the predicted value as a sample, so that the aim of optimizing the neural network is not achieved; therefore, after the predicted value is added into the sample set and the neural network is trained to acquire dining data information of a plurality of days in the future, the neural network model before prediction needs to be recovered; the method for recovering the neural network model before prediction comprises the following steps: before predicting restaurant data information for a plurality of days in the future, storing check points of the model, and storing current model parameters and weights on a disk; and loading the previously stored model immediately after the catering data information of a plurality of days in the future is predicted, so as to restore the neural network model before prediction.
In addition, the invention also provides a restaurant data modeling system based on the intelligent marketing scene, which can realize the method, and comprises the following steps: the system comprises a data acquisition unit, a data processing unit, a data analysis unit and a model updating unit; the data acquisition unit is connected with the front-end equipment of the merchant and the data processing unit and is used for acquiring restaurant data information from the front-end equipment of the merchant and sending the acquired restaurant data information to the data processing unit; the data processing unit is connected with the data acquisition unit and the data analysis unit and is used for processing the data sent by the data acquisition unit and sending the processed data to the data analysis unit; the data analysis unit is connected with the data processing unit and the model updating unit and is used for analyzing the data sent by the data processing unit and sending the analyzed neural network model to the model updating unit; the model updating unit is connected with the data analysis unit and is used for updating the neural network model trained by the acquired data analysis unit and predicting restaurant data information.
Specifically: the data acquisition unit acquires catering data information through merchant front-end equipment, classifies the catering data information, and divides the catering data information into first class data information, second class data information and third class data information; the data processing unit performs normalization processing on the received first type data information and second type data information, and extracts the first type data information and the second type data information respectivelyCharacteristic information data, and a catering data information sample set formed by combining the extracted characteristic information data and third type data information is { (X) i ,Y i ) -a }; the data analysis unit obtains a data information sample set { (X) i ,Y i ) And collect the sample { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain a trained neural network model; the model updating unit can predict catering data information of a first day in the future and a plurality of days in the future; meanwhile, after the catering data information of the first day in the future is predicted, the neural network model before prediction is optimized through the catering data information of the actual day different from the predicted catering data information, and the neural network model is continuously optimized, so that the neural network model is more in line with the actual operation mode of the catering industry, and the situation that the predicted value is more and more inaccurate due to the change of the mode of the catering industry and the change of the dining habit of a customer is avoided.
The beneficial effects of the invention are as follows:
the invention provides a restaurant data modeling method and a restaurant data modeling system based on an intelligent marketing scene, wherein the method firstly divides restaurant data information into three types, respectively normalizes first-class data information and second-class data information, and obtains characteristic information data of the first-class data information and the second-class data information; combining the obtained characteristic information data with the third type of data information to form a restaurant data information sample set { (X) i ,Y i ) -a }; then training the catering data information sample set through an LSTM neural network to obtain a neural network model; finally, optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, continuously optimizing the neural network model, and predicting the catering data information of a plurality of days in the future through the optimized neural network model; the invention divides the complex food and beverage data information into three categories, which is convenient for processing the complex food and beverage data information; training the catering data information through an LSTM neural network to obtain a trained neural network model; finally, a method for updating and optimizing the neural network model is providedThe updating and optimizing method enables the neural network model to be adjusted along with the change of the catering industry mode, and avoids the situation that the predicted value of the neural network model is more and more inaccurate due to the change of the catering industry mode and the change of the dining habits of clients.
Drawings
Fig. 1: the invention discloses a catering data modeling method based on an intelligent marketing scene.
Fig. 2: the invention discloses a schematic diagram of a catering data modeling system based on an intelligent marketing scene.
Detailed Description
Specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the particular embodiments presented herein are illustrative and explanatory only and are not restrictive of the invention.
It should be noted that numerous specific details are set forth in the following description in order to provide a thorough understanding of the present invention, however, that other embodiments of the invention and variations thereof are possible and, therefore, the scope of the invention is not limited by the specific examples disclosed below.
As shown in fig. 1, a restaurant data modeling method based on an intelligent marketing scenario according to an embodiment of the present invention includes: step S100, acquiring restaurant data information in the last three years; the dining data information comprises: first class data information, second class data information and third class data information; the first type of data information is merchant dish sales data information, and comprises: the total quantity information of dishes and the corresponding quantity of each dish variety; the second type of data information is client dining data information, and comprises the following steps: old customers and new customers have dinner times; the third type of data information is profit information corresponding to the first type of data information and the second type of data information in a period of time; step S101, respectively carrying out normalization processing on first-class data information and second-class data information by taking each day as a unit to obtain characteristic information data of the first-class data information and the second-class data information; the characteristic information data of the first type data information and the second type data information after normalization processing is combined with the first type dataThe three kinds of data information are combined to form a catering data information sample set { (X) i ,Y i ) Wherein X is i Representing a vector composed of characteristic information data of the first type of data information and the second type of data information, Y i Representation and vector X i Corresponding profit information; step S102, obtaining a food and beverage data information sample set { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain trained neural network model, wherein X i As an input sequence to the LSTM neural network, Y i A target sequence corresponding to the LSTM neural network; step S103, predicting the catering data information of the first day in the future through a trained neural network model, and acquiring the catering data information of the actual day according to the management condition of the actual day; then, comparing the predicted dining data information with the dining data information of the actual day, if the predicted dining data information is different from the dining data information of the actual day; optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, and continuously optimizing the neural network model; step S104, predicting dining data information of a plurality of days in the future through an optimized neural network model; the prediction method comprises the following steps: firstly, predicting restaurant data information of a first day in the future through an optimized neural network model; then, predicting the catering data information of the second future day by taking the predicted catering data information of the first future day as the data of the training set; obtaining dining data information of multiple days in the future through repeated prediction; and after the catering data information of a plurality of days in the future is acquired, recovering the neural network model before prediction.
Specifically, the method comprises the steps of firstly dividing restaurant data information into first-class data information, second-class data information and third-class data information, respectively carrying out normalization processing on the first-class data information and the second-class data information, and obtaining characteristic information data of the first-class data information and the second-class data information; combining the obtained characteristic information data with the third type of data information to form a restaurant data information sample set { (X) i ,Y i ) -a }; subsequently, the restaurant data information is sampledThe set is trained through an LSTM neural network to obtain a neural network model; and finally, optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, continuously optimizing the neural network model, and predicting the catering data information of a plurality of days in the future through the optimized neural network model.
Step S100, acquiring restaurant data information in the last three years; the dining data information comprises: first class data information, second class data information and third class data information; the first type of data information is merchant dish sales data information, and comprises: the total quantity information of dishes and the corresponding quantity of each dish variety; the second type of data information is client dining data information, and comprises the following steps: old customers and new customers have dinner times; the third type of data information is profit information corresponding to the first type of data information and the second type of data information in a period of time; specifically, according to the actual conditions of the catering company, the catering data information in the last three years is obtained, and the catering data information is classified into: first class data information, second class data information and third class data information.
In the above embodiment, preferably, in the case of acquiring data information, in recent years, the invention selects food and beverage data information of about three years through the food and beverage information database; because the catering industry mode is continuously changed, if the catering data information of earlier time is selected, the situation that the training model has larger difference between the prediction result of the current catering data information and the actual situation is caused because the catering industry mode of earlier time does not accord with the current operation situation.
Step S101, respectively carrying out normalization processing on first-class data information and second-class data information by taking each day as a unit to obtain characteristic information data of the first-class data information and the second-class data information; combining the characteristic information data of the normalized first type data information and the normalized second type data information with the third type data information to form a restaurant data information sample set { (X) i ,Y i ) Wherein X is i Representing the information of the first type and the second type of dataVector of characteristic information data of information, Y i Representation and vector X i Corresponding profit information; products related to the catering industry are numerous, so that the first type data information, the second type data information and the third type data information which are directly acquired cannot be directly converted, and the relationship between the first type data information, the second type data information and the third type data information is found; and because the third type of data information is profit information, only the first type of data information and the second type of data information are normalized.
In the above embodiment, specifically, the first type data information and the second type data information are converted into feature information data that is easy to compare by adopting a feature scaling normalization processing manner; the characteristic data information of the first type data information and the second type data information together form a data set { X } i (wherein X is i A vector formed by characteristic information data of the first type data information and the second type data information; regarding the data set composed of the third type of data information as { Y } i },Y i Corresponding profit information; the catering data information sample set consisting of the first type of data information, the second type of data information and the third type of data information is { (X) i ,Y i )}。
Step S102, obtaining a food and beverage data information sample set { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain trained neural network model, wherein X i As an input sequence to the LSTM neural network, Y i A target sequence corresponding to the LSTM neural network; specifically, the input sequence is an LSTM neural network input end input sample sequence; the target sequence is used for adjusting the output end data of the LSTM neural network, so that the output end data of the LSTM neural network is closer to the target sequence.
In the above embodiment, specifically, the obtained restaurant data information sample set { (X) i ,Y i ) The training process of training by inputting the LSTM neural network is as follows: a) Sample set { (X) i ,Y i ) Inputting into LSTM neural network to obtain model output; b) Comparing the output of the model with the target sequence, and calculating a loss function value; c) Computing gradients using back propagation and using optimizersUpdating the model parameters to reduce the value of the loss function; d) Repeating the steps until reaching the specified training condition, stopping training and obtaining the neural network model.
In the above embodiments, in particular, the LSTM neural network is a variant of a Recurrent Neural Network (RNN) with the ability to memorize long and short-term information; the LSTM neural network is used for processing the catering data information, so that the characteristic information of the catering data information at different times can be learned, and the characteristic information can be used for predicting future catering data information; when restaurant data information is processed through the LSTM neural network, the LSTM model is set as follows: a) The time step of the LSTM model is set to 7, the input dimension is set to 4, and the output dimension is set to 1; b) The activating function of the LSTM layer is a ReLU function, and the activating function of the output layer is a sigmoid function; c) Setting the first, second and third LSTM hidden layer neurons as 256, 256 and 128 respectively; meanwhile, the hidden layer neurons are deleted randomly by adopting Dropout to prevent overfitting, and the random inactivation rate is set to be 0.3; d) Training an LSTM neural network model through tensorfilow, wherein a loss function of the model uses MSE, the learning rate of the model is set to be 0.01, and the initial value of the bias parameter is set to be 0; the batch size, epoch of the model were set to 100 and 200, respectively.
In the above embodiment, specifically, the method until reaching the specified condition of training: when the fixed epoch value is reached, the verification loss is not reduced, the verification performance is not improved, the performance reaches the target and the time limit, and the like, training is stopped, and the neural network model is successfully generated.
Step S103, predicting the catering data information of the first day in the future through a trained neural network model, and acquiring the catering data information of the actual day according to the management condition of the actual day; then, comparing the predicted dining data information with the dining data information of the actual day, if the predicted dining data information is different from the dining data information of the actual day; optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, and continuously optimizing the neural network model; specifically, inputting a date, and predicting dining data information of the first day in the future through a trained neural network model; according to the input date and the actual business condition of the date and the day, acquiring the catering data information of the actual day; comparing the predicted dining data information with the dining data information of the actual day, if the predicted dining data information is found to be different from the dining data information of the actual day, the predicted dining data information is caused by accidental events or changes of modes of the dining industry, so that the neural network model before prediction is optimized through the dining data information of the actual day which is different from the predicted dining data information; if the predicted dining data information is found to be the same as the dining data information of the actual day, the neural network model is not performed; the reason for not optimizing is that the predicted data is generated from the model, the predicted dining data information is the same as the dining data information of the actual day, only the model prediction can be illustrated to be more accurate, and the predicted data is the embodiment of the neural network model; therefore, if the neural network model before prediction is trained through the catering data information of the actual day which is the same as the predicted catering data information, the optimization effect cannot be achieved; the method is also a highlight point, and only the catering data information of the actual day different from the predicted catering data information is adopted to optimize the neural network model before prediction, so that the optimization times and calculation resources are saved.
In the foregoing embodiment, specifically, the predicted dining data information is a prediction of the dining data information of the first day in the future by the trained neural network model, and the dining data information of the actual day refers to acquiring, according to the actual business situation, the dining data information of the actual day consistent with the predicted date; if the predicted dining data information and the dining data information of the actual day are different, the following descriptions are provided: instability of restaurant data information, insufficient generalization capability of a model and unknown future information; meanwhile, the neural network model before prediction is optimized through the catering data information of the actual day different from the predicted catering data information, so that the stability of the neural network model for prediction of the catering industry can be increased, and the optimization of the neural network model is realized.
Step S104, predicting dining data information of a plurality of days in the future through an optimized neural network model; the prediction method comprises the following steps: firstly, predicting restaurant data information of a first day in the future through an optimized neural network model; then, predicting the catering data information of the second future day by taking the predicted catering data information of the first future day as the data of the training set; obtaining dining data information of multiple days in the future through repeated prediction; after acquiring dining data information of a plurality of days in the future, recovering a neural network model before prediction; specifically, along with the continuous optimization of the neural network model in step S103, the dining data information of multiple days in the future is predicted by the optimized neural network model.
In the above embodiment, specifically, it should be noted that, by taking the predicted value as a sample set, training is performed on the neural network, and future dining data information is predicted; the predicted value is not catering data information of the actual day, but is a data embodiment of the neural network model; training the neural network model by taking the predicted value as a sample, so that the aim of optimizing the neural network model is not achieved; therefore, after the predicted value is added into the sample set and the neural network model is trained, the neural network model in the first future prediction day is needed to be recovered after the catering data information of a plurality of future days is obtained; the method for recovering the neural network model before prediction comprises the following steps: before predicting restaurant data information for a plurality of days in the future, storing check points of a neural network model, and storing current model parameters and weights on a disk; and loading the previously stored model immediately after the catering data information of a plurality of days in the future is predicted, so as to restore the neural network model before prediction.
As shown in fig. 2: the invention relates to a schematic diagram of a catering data modeling system based on an intelligent marketing scene; the figure includes: s200, a data acquisition unit; s201, a data processing unit; s202, a data analysis unit; s203, a model updating unit.
In the above embodiment, specifically, the data acquisition unit S200 is provided with a data acquisition tool, which is used for acquiring restaurant data information on the front-end device of the merchant, and sending the acquired restaurant data information to the data processing unit S201; the data processing unit S201 is installed with data processing software such as: numPy and Pandas library of Python, R language and various packages thereof can normalize the received restaurant data information; the data analysis unit S202 is installed with related software loaded with LSTM neural networks, such as: tensorFlow, keras and PyTorch, etc., can realize the construction, training, optimization and application of LSTM neural network; the model updating unit S203 is loaded with the same LSTM neural network related software as S202, but installs a storage space supported by the disk at the same time, so as to save checkpoints of the neural network model, and save current model parameters and weights to the disk, so as to facilitate recovery of the neural network model before prediction.
In the above embodiment, specifically, the data acquisition unit S200 is connected to the front end device of the merchant and the data processing unit S201, and is configured to acquire restaurant data information from the front end device of the merchant, and send the acquired restaurant data information to the data processing unit S201; the data processing unit S201 is connected with the data acquisition unit S200 and the data analysis unit S202, and is used for processing the data sent by the data acquisition unit S200 and sending the processed data to the data analysis unit S202; the data analysis unit S202 is connected with the data processing unit S201 and the model updating unit S203, and is used for analyzing the data sent by the data processing unit S201 and sending the analyzed neural network model to the model updating unit S203; the model updating unit S203 is connected to the data analysis unit S202, and is configured to update the neural network model trained by the acquired data analysis unit S202, and predict dining data information.
In the above embodiment, specifically, the data acquisition unit S200 acquires the catering data information through the front-end device of the merchant, classifies the catering data information into the first type of data information, the second type of data information and the third type of data information; the data processing unit S201 performs normalization processing on the received first-class data information and second-class data information, and extracts the characteristic information numbers of the first-class data information and the second-class data information respectivelyAccording to the food and beverage data information sample set formed by combining the extracted characteristic information data and the third type of data information is { (X) i ,Y i ) -a }; the data analysis unit S202 obtains a data information sample set { (X) i ,Y i ) And collect the sample { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain a trained neural network model; the model updating unit S203 is capable of predicting dining data information of a future first day and a future plurality of days; meanwhile, after the catering data information of the first day in the future is predicted, the neural network model before prediction is optimized through the catering data information of the actual day different from the predicted catering data information, so that the aim of continuously optimizing the neural network model is fulfilled; the neural network model is more in line with the actual operation mode of the catering industry, and the situation that the predicted value is more and more inaccurate due to the change of the mode of the catering industry and the change of the dining habits of clients is avoided.
It is to be understood that the above-described embodiments are one or more embodiments of the invention, and that many other embodiments and variations thereof are possible in accordance with the invention; variations and modifications of the invention, which are intended to be within the scope of the invention, will occur to those skilled in the art without any development of the invention.
Claims (9)
1. A restaurant data modeling method based on an intelligent marketing scenario, the method comprising:
1) Collecting catering data information in the last three years; the dining data information comprises: first class data information, second class data information and third class data information; the first type of data information is merchant dish sales data information, and comprises: the total quantity information of dishes and the corresponding quantity of each dish variety; the second type of data information is client dining data information, and comprises the following steps: old customers and new customers have dinner times; the third type of data information is profit information corresponding to the first type of data information and the second type of data information in a period of time;
2) For the first type of data information and the first type of data information respectively in each day unitNormalizing the second-class data information to obtain characteristic information data of the first-class data information and the second-class data information; combining the characteristic information data of the normalized first type data information and the normalized second type data information with the third type data information to form a restaurant data information sample set { (X) i ,Y i ) Wherein X is i Representing a vector composed of characteristic information data of the first type of data information and the second type of data information, Y i Representation and vector X i Corresponding profit information;
3) The obtained restaurant data information sample set { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain trained neural network model, wherein X i As an input sequence to the LSTM neural network, Y i A target sequence corresponding to the LSTM neural network; the food and beverage data information sample set to be obtained { (X) i ,Y i ) The training process of training by inputting the LSTM neural network is as follows: a) Sample set { (X) i ,Y i ) Inputting into LSTM neural network to obtain model output; b) Comparing the output of the model with the target sequence, and calculating a loss function value; c) Calculating gradients using a back propagation method and updating model parameters with an optimizer to reduce the value of the loss function; d) Repeating the steps until reaching the specified training condition, stopping training and obtaining a neural network model;
4) Predicting the catering data information of the first day in the future through a trained neural network model, and acquiring the catering data information of the actual day according to the management condition of the actual day; then, comparing the predicted dining data information with the dining data information of the actual day, if the predicted dining data information is different from the dining data information of the actual day; optimizing the neural network model before prediction through the catering data information of the actual day different from the predicted catering data information, and continuously optimizing the neural network model;
5) Predicting dining data information of a plurality of days in the future through an optimized neural network model; the prediction method comprises the following steps: firstly, predicting restaurant data information of a first day in the future through an optimized neural network model; then, predicting the catering data information of the second future day by taking the predicted catering data information of the first future day as the data of the training set; obtaining dining data information of multiple days in the future through repeated prediction; and after the catering data information of a plurality of days in the future is acquired, recovering the neural network model before prediction.
2. The method for modeling restaurant data based on intelligent marketing scenarios of claim 1, wherein the collecting restaurant data information over the last three years comprises: according to the actual operating conditions of catering companies, acquiring catering data information in the last three years, classifying the catering data information, and dividing the catering data information into: first class data information, second class data information and third class data information.
3. The method for modeling restaurant data based on intelligent marketing scenario of claim 1, wherein normalizing the first type of data information and the second type of data information, respectively, comprises: converting the first type data information and the second type data information into characteristic information data which is easy to compare by adopting a characteristic scaling normalization processing mode; the characteristic data information of the first type data information and the second type data information together form a data set { X } i (wherein X is i A vector formed by characteristic information data of the first type data information and the second type data information; regarding the data set composed of the third type of data information as { Y } i },Y i Corresponding profit information; the catering data information sample set consisting of the first type of data information, the second type of data information and the third type of data information is { (X) i ,Y i )}。
4. The method for modeling restaurant data based on intelligent marketing scenarios according to claim 1, wherein the sample set of restaurant data information is { (X) i ,Y i ) Training the input LSTM neural network includes: the LSTM neural network is a variant of a Recurrent Neural Network (RNN) withThe memory device has the capability of memorizing long and short time information; the LSTM neural network is used for processing the catering data information, so that the characteristic information of the catering data information at different times can be learned, and the characteristic information can be used for predicting future catering data information; when restaurant data information is processed through the LSTM neural network, the LSTM model is set as follows: a) The time step of the LSTM model is set to 7, the input dimension is set to 4, and the output dimension is set to 1; b) The activating function of the LSTM layer is a ReLU function, and the activating function of the output layer is a sigmoid function; c) Setting the first, second and third LSTM hidden layer neurons as 256, 256 and 128 respectively; meanwhile, the hidden layer neurons are deleted randomly by adopting Dropout to prevent overfitting, and the random inactivation rate is set to be 0.3; d) Training an LSTM neural network model through tens orflow, wherein a loss function of the model uses MSE, the learning rate of the model is set to be 0.01, and the initial value of the bias parameter is set to be 0; the batch size, epoch of the model were set to 100 and 200, respectively.
5. The method for modeling restaurant data based on intelligent marketing scenarios of claim 1, wherein the specified conditions until training is reached include: the specified conditions are conditions for the LSTM neural network to reach a stop training, including: achieving a fixed epoch value, no further reduction in validation loss, no further improvement in validation performance, performance meeting target and time constraints.
6. The method for modeling dining data based on intelligent marketing scenarios of claim 1, wherein the if the predicted dining data information is different from the dining data information of the actual day comprises: the predicted dining data information is the prediction of the trained neural network model on dining data information of the first day in the future, and the dining data information of the actual day refers to the acquisition of the dining data information of the actual day consistent with the predicted date according to the actual business situation; if the predicted dining data information and the dining data information of the actual day are different, the following descriptions are provided: instability of restaurant data information, insufficient generalization capability of a model and unknown future information; meanwhile, the neural network model before prediction is optimized through the catering data information of the actual day different from the predicted catering data information, so that the stability of the neural network model for prediction of the catering industry can be increased, and the optimization of the neural network model is realized.
7. The method for modeling dining data based on intelligent marketing scenarios of claim 1, wherein recovering the pre-forecast neural network model after obtaining the dining data information for a plurality of days in the future comprises: training the neural network model by taking the predicted value as a sample set, and predicting future catering data information; the predicted value is not catering data information of the actual day, but is a data embodiment of the neural network model; training the neural network model by taking the predicted value as a sample, so that the aim of optimizing the neural network model is not achieved; therefore, after the predicted value is added into the sample set and the neural network model is trained, the neural network model before prediction needs to be recovered after the catering data information of a plurality of days in the future is obtained; the method for recovering the neural network model before prediction comprises the following steps: before predicting restaurant data information for a plurality of days in the future, storing check points of the model, and storing current model parameters and weights on a disk; and loading the previously stored model immediately after the catering data information of a plurality of days in the future is predicted, so as to restore the neural network model before prediction.
8. A restaurant data modeling system based on intelligent marketing scenarios, the system comprising: the system comprises a data acquisition unit, a data processing unit, a data analysis unit and a model updating unit; the data acquisition unit is connected with the front-end equipment of the merchant and the data processing unit and is used for acquiring restaurant data information from the front-end equipment of the merchant and sending the acquired restaurant data information to the data processing unit; the data processing unit is connected with the data acquisition unit and the data analysis unit and is used for processing the data sent by the data acquisition unit and sending the processed data to the data analysis unit; the data analysis unit is connected with the data processing unit and the model updating unit and is used for analyzing the data sent by the data processing unit and sending the analyzed neural network model to the model updating unit; the model updating unit is connected with the data analysis unit and is used for updating the neural network model trained by the acquired data analysis unit and predicting restaurant data information.
9. The smart marketing scenario-based dining data modeling system of claim 8, wherein the system further comprises: the data acquisition unit acquires catering data information through merchant front-end equipment, classifies the catering data information, and divides the catering data information into first class data information, second class data information and third class data information; the data processing unit performs normalization processing on the received first type data information and second type data information, extracts characteristic information data of the first type data information and the second type data information respectively, and combines the extracted characteristic information data and third type data information to form a catering data information sample set as { (X) i ,Y i ) -a }; the data analysis unit obtains a data information sample set { (X) i ,Y i ) And collect the sample { (X) i ,Y i ) Inputting into LSTM neural network for training to obtain a trained neural network model; the model updating unit can predict catering data information of a first day in the future and a plurality of days in the future; meanwhile, after the catering data information of the first day in the future is predicted, the neural network model before prediction is optimized through the catering data information of the actual day different from the predicted catering data information, and the neural network model is continuously optimized, so that the neural network model is more in line with the actual operation mode of the catering industry, and the situation that the predicted value is more and more inaccurate due to the change of the mode of the catering industry and the change of the dining habit of a customer is avoided.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311547883.3A CN117557289B (en) | 2023-11-20 | 2023-11-20 | Restaurant data modeling method and system based on intelligent marketing scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311547883.3A CN117557289B (en) | 2023-11-20 | 2023-11-20 | Restaurant data modeling method and system based on intelligent marketing scene |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117557289A true CN117557289A (en) | 2024-02-13 |
CN117557289B CN117557289B (en) | 2024-04-16 |
Family
ID=89816283
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311547883.3A Active CN117557289B (en) | 2023-11-20 | 2023-11-20 | Restaurant data modeling method and system based on intelligent marketing scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117557289B (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292672A (en) * | 2017-07-05 | 2017-10-24 | 上海数道信息科技有限公司 | System and method for is realized in a kind of catering industry sales forecast |
CN108364092A (en) * | 2018-01-29 | 2018-08-03 | 西安理工大学 | A kind of catering trade vegetable Method for Sales Forecast method based on deep learning |
US20180276691A1 (en) * | 2017-03-21 | 2018-09-27 | Adobe Systems Incorporated | Metric Forecasting Employing a Similarity Determination in a Digital Medium Environment |
WO2020143409A1 (en) * | 2019-01-07 | 2020-07-16 | 阿里巴巴集团控股有限公司 | Method and device for predicting business indicators |
CN111626764A (en) * | 2020-04-09 | 2020-09-04 | 中南大学 | Commodity sales volume prediction method and device based on Transformer + LSTM neural network model |
CN111667304A (en) * | 2020-05-21 | 2020-09-15 | 北京智通云联科技有限公司 | Method and system for predicting product sales under multi-mode condition |
CN114418071A (en) * | 2022-01-24 | 2022-04-29 | 中国光大银行股份有限公司 | Cyclic neural network training method |
CN115994785A (en) * | 2023-01-09 | 2023-04-21 | 淮阴工学院 | Intelligent prediction method and system for catering traffic stock |
CN116205338A (en) * | 2022-12-30 | 2023-06-02 | 浪潮工业互联网股份有限公司 | Method, device, equipment and medium for predicting dish sales |
WO2023120126A1 (en) * | 2021-12-20 | 2023-06-29 | 株式会社日立製作所 | Unit-sales prediction system and unit-sales prediction method |
KR102576104B1 (en) * | 2022-06-02 | 2023-09-06 | 임현진 | System for providing consulting service for restaurant |
-
2023
- 2023-11-20 CN CN202311547883.3A patent/CN117557289B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180276691A1 (en) * | 2017-03-21 | 2018-09-27 | Adobe Systems Incorporated | Metric Forecasting Employing a Similarity Determination in a Digital Medium Environment |
CN107292672A (en) * | 2017-07-05 | 2017-10-24 | 上海数道信息科技有限公司 | System and method for is realized in a kind of catering industry sales forecast |
CN108364092A (en) * | 2018-01-29 | 2018-08-03 | 西安理工大学 | A kind of catering trade vegetable Method for Sales Forecast method based on deep learning |
WO2020143409A1 (en) * | 2019-01-07 | 2020-07-16 | 阿里巴巴集团控股有限公司 | Method and device for predicting business indicators |
CN111626764A (en) * | 2020-04-09 | 2020-09-04 | 中南大学 | Commodity sales volume prediction method and device based on Transformer + LSTM neural network model |
CN111667304A (en) * | 2020-05-21 | 2020-09-15 | 北京智通云联科技有限公司 | Method and system for predicting product sales under multi-mode condition |
WO2023120126A1 (en) * | 2021-12-20 | 2023-06-29 | 株式会社日立製作所 | Unit-sales prediction system and unit-sales prediction method |
CN114418071A (en) * | 2022-01-24 | 2022-04-29 | 中国光大银行股份有限公司 | Cyclic neural network training method |
KR102576104B1 (en) * | 2022-06-02 | 2023-09-06 | 임현진 | System for providing consulting service for restaurant |
CN116205338A (en) * | 2022-12-30 | 2023-06-02 | 浪潮工业互联网股份有限公司 | Method, device, equipment and medium for predicting dish sales |
CN115994785A (en) * | 2023-01-09 | 2023-04-21 | 淮阴工学院 | Intelligent prediction method and system for catering traffic stock |
Also Published As
Publication number | Publication date |
---|---|
CN117557289B (en) | 2024-04-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Vanderveld et al. | An engagement-based customer lifetime value system for e-commerce | |
US8650079B2 (en) | Promotion planning system | |
CN111160968A (en) | SKU-level commodity sales prediction method and device | |
US9449281B2 (en) | Statistical machine learning | |
US20180349790A1 (en) | Time-Based Features and Moving Windows Sampling For Machine Learning | |
US7664671B2 (en) | Methods and systems for profile-based forecasting with dynamic profile selection | |
EP1723596A1 (en) | System for individualized customer interaction | |
US20210125073A1 (en) | Method and system for individual demand forecasting | |
US20220398610A1 (en) | Method of forecasting store demand based on artificial intelligence and system therefor | |
US20220108334A1 (en) | Inferring unobserved event probabilities | |
Koumétio et al. | Optimizing the prediction of telemarketing target calls by a classification technique | |
CN113469598A (en) | Intelligent supply chain system and server platform | |
WO2023225529A2 (en) | Predictive systems and processes for product attribute research and development | |
WO2021077226A1 (en) | Method and system for individual demand forecasting | |
Andrade et al. | Disaggregated retail forecasting: A gradient boosting approach | |
CN117557289B (en) | Restaurant data modeling method and system based on intelligent marketing scene | |
US20200193486A1 (en) | System and method for determining bid vector transformed predictive click-through rate | |
CA3059904A1 (en) | Method and system for generating aspects associated with a future event for a subject | |
Ehsani | Customer churn prediction from Internet banking transactions data using an ensemble meta-classifier algorithm | |
CN116167646A (en) | Evaluation method, device, equipment and storage medium based on transaction algorithm | |
US20220327594A1 (en) | Systems and methods for useful life determination of target items | |
WO2021077227A1 (en) | Method and system for generating aspects associated with a future event for a subject | |
US12020124B2 (en) | Selecting optimum primary and secondary parameters to calibrate and generate an unbiased forecasting model | |
CN113240359A (en) | Demand prediction method for coping with external serious fluctuation | |
US20210125031A1 (en) | Method and system for generating aspects associated with a future event for a subject |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |