Canteen satisfaction quantitative evaluation method, electronic equipment and medium

Abstract

The invention provides a canteen satisfaction quantitative evaluation method, electronic equipment and a medium. The method comprises the following steps: acquiring Internet of Things sensing data and sales volume of each dish; obtaining dining scene data based on the monitoring video stream of the canteen; obtaining an active feedback text of the diner; feedback information is extracted from the active feedback text; according to the data acquisition time, the dish name and the window information, data fusion is carried out on the Internet of Things sensing data, the sales volume, the dining scene data and the feedback information to obtain a multi-source data association model, and the multi-source data association model comprises multiple groups of association data; determining a dish taste index of each dish, a service efficiency index of each window and an environment comfort level of a canteen based on the multiple groups of associated data; and calculating the weighted sum of the dish taste index, the service efficiency index and the environmental comfort to obtain comprehensive satisfaction. According to the scheme, the satisfaction degree of the canteen can be quantified more accurately.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of canteen satisfaction evaluation, and more particularly to a canteen satisfaction quantitative evaluation method, an electronic device and a medium. BACKGROUND

[0002] As the core place for providing centralized dining services for teachers and students or employees in schools, enterprises and other units, canteens play an important role in daily work and life. The canteen satisfaction evaluation is a key link for accurately optimizing canteen operation management and improving service quality.

[0003] In related technologies, the canteen satisfaction evaluation usually relies on questionnaire survey and active scoring by diners. This evaluation method has small sample size, strong subjectivity and survivorship bias (only extremely satisfied or dissatisfied diners are willing to provide feedback), and cannot represent the real opinions of most silent diners.

[0004] Therefore, the present application is proposed. SUMMARY

[0005] The present application is proposed in consideration of the above problems. According to one aspect of the present application, a canteen satisfaction quantitative evaluation method is provided, comprising: obtaining Internet of Things (IoT) sensing data and sales volume of each dish, wherein the IoT sensing data includes initial weight and remaining amount of the corresponding dish; obtaining dining scene data based on a monitoring video stream of the canteen, wherein the dining scene data includes queue detection data, regional personnel density data and seat occupancy rate of the canteen; obtaining active feedback text of diners; extracting feedback information from the active feedback text, wherein the feedback information includes dish score, dish evaluation, comfort degree score and comfort degree evaluation of the canteen; performing data fusion on the IoT sensing data, the sales volume, the dining scene data and the feedback information according to data collection time, dish name and window information, to obtain a multi-source data association model, wherein the multi-source data association model includes multiple sets of associated data; determining dish taste index of each dish, service efficiency index of each window and environmental comfort degree of the canteen based on the multiple sets of associated data; calculating weighted sum of the dish taste index, the service efficiency index and the environmental comfort degree, to obtain a comprehensive satisfaction degree.

[0006] Exemplarily, the determination of the dish taste index of each dish comprises: for any dish, determining a remaining meal rate of the dish based on the initial weight and the remaining amount of the dish. determining an average score of the dish based on dish scores of the dishes on the dish; determining a dish taste index of the dish based on the leftover rate, the average score, and a sales volume of the dish.

[0007] Exemplarily, the method further comprises: for each dish, determining whether the leftover rate of the dish is greater than a first leftover rate threshold; determining whether the dish taste index of the dish is less than a taste index threshold; when the leftover rate of the dish is greater than the first leftover rate threshold and the dish taste index of the dish is less than the taste index threshold, obtaining a negative dish evaluation on the dish in the feedback information; combining the leftover rate, the dish taste index, and the negative dish evaluation of the dish into a transaction item set and outputting.

[0008] Exemplarily, the method further comprises: predicting a demand volume of a dish corresponding to a next meal period based on a leftover rate and a sales volume of the dish.

[0009] Exemplarily, the method further comprises: for each dish, when the leftover rate of the dish is greater than a second leftover rate threshold and the sales volume of the dish is less than a historical average sales volume of the dish, generating a dish slow-selling warning.

[0010] Exemplarily, the determining of the service efficiency index of each window comprises: for any window, determining an average queuing time length of the window based on queue detection data corresponding to the window; determining the service efficiency index based on the average queuing time length.

[0011] Exemplarily, the determining of the environmental comfort degree of the canteen comprises: determining an average comfort degree score based on comfort degree scores of each diner; counting numbers of positive feedback keywords and negative feedback keywords in the comfort degree evaluations; determining a positive feedback ratio and a negative feedback ratio based on the numbers of the positive feedback keywords and the negative feedback keywords; determining the environmental comfort degree based on the area person density data, the seat occupancy rate, the average comfort degree score, the positive feedback ratio, and the negative feedback ratio.

[0012] Exemplarily, the method further comprises: The service efficiency index obtained each time is taken as a dependent variable, and the corresponding average queuing time, seat occupancy rate and working window number are taken as independent variables to fit to obtain a regression equation; Regression analysis is performed based on the regression equation to determine the influence degree of each independent variable on the service efficiency index.

[0013] According to still another aspect of the present application, an electronic device is provided, which includes a processor and a memory having a computer program stored therein, the processor being configured to execute the computer program to implement the method as described above.

[0014] According to yet another aspect of the present application, a computer readable storage medium is provided, which stores a computer program / instruction, the computer program / instruction being configured to implement the method as described above when executed by a processor.

[0015] In the above technical solution, the satisfaction degree of the canteen is evaluated based on Internet of Things sensing data, sales and monitoring video streams of each dish and active feedback text of diners, wherein the multi-source passive data can reflect the real behavior of a large number of diners, and the active feedback text can reflect the subjective feelings, core demands and potential pain points of diners. On the one hand, the data from different sources and of different types are associated and integrated, which is beneficial to breaking the island and forming a unified satisfaction analysis view. On the other hand, the combination of the two can more objectively and comprehensively evaluate the satisfaction degree of the canteen, which can not only avoid subjective bias caused by the one-sided words of a small number of people, improve the scientificity and accuracy of the results, and on the other hand, the scheme can automatically generate the satisfaction degree of the canteen based on multi-source data, greatly reducing the manual intervention. Greatly improve the management efficiency and reduce the labor cost. And this method can dynamically monitor based on real-time data or quasi-real-time data stream, so as to output the comprehensive satisfaction degree in real time based on the current data, the feedback cycle is short, which helps to provide more accurate decision basis for canteen management optimization, realizes the fundamental change from "experience driven" to "data driven", "lagging rectification" to "active prediction", "fuzzy perception" to "accurate quantification", completely changes the dilemma of traditional canteen satisfaction management relying on subjective, lagging and one-sided information, and quickly responds to the needs of diners. In addition, this way builds a multi-dimensional quantitative index system covering dish quality (represented by dish taste index), service quality (represented by service efficiency index), and environment experience (represented by environment comfort), each index is supported by corresponding data source. This multi-dimensional and refined evaluation method can further improve the scientificity and accuracy of the results, and provide strong core support for the fine and intelligent operation of the canteen. It is especially suitable for scenes such as courts with fixed personnel structure, peak concentration and high privacy requirements. It can provide a more objective, real-time, comprehensive and in-depth satisfaction quantitative management and decision support tool on the basis of meeting the management specifications and security disciplines of government agencies, effectively improving the dining experience of diners and the management effect of the canteen, optimizing the operation cost and management efficiency of the canteen, and perfectly meeting the urgent needs of intelligent upgrading of the logistics support system in the construction of smart courts.

[0016] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the contents of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0017] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which like reference characters refer to like parts throughout the figures. The drawings provided are for illustrative purposes only and, therefore, should not be considered to limit the present application. In the drawings:

[0018] Figure 1 A canteen satisfaction quantification evaluation method according to one embodiment of the present application is shown; Figure 2 A schematic block diagram of an electronic device according to one embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] In order to make the objects, technical solutions and advantages of the present application more obvious, the following will describe the example embodiments according to the present application in detail with reference to the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application described in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the protection scope of the present application.

[0020] As described above, the current canteen satisfaction evaluation method usually relies on questionnaire survey and active scoring by diners, which has small sample size, strong subjectivity, and "survivor bias" (only extremely satisfied or dissatisfied diners are willing to feedback), and cannot represent the real opinions of most silent diners. As a result, the feedback is distorted, which not only cannot effectively guide the canteen management to improve, but also may mislead the canteen management and reduce the quality of canteen management. Moreover, the traditional questionnaire analysis has long cycle and feedback lag, and when the management party finds the problem, the problem may have lasted for several days, causing negative impact. In view of this, the present application provides a canteen satisfaction quantification evaluation method, an electronic device and a medium, which evaluates the canteen satisfaction based on the data stream obtained by fusing multi-source passive data (Internet of Things sensing data and sales of each dish, and monitoring video stream of the canteen) and active data (active feedback text), can more objectively and comprehensively evaluate the canteen satisfaction, reduce subjective bias, and the method can be dynamically monitored based on real-time data or quasi-real-time data stream, so that the comprehensive satisfaction can be output in real time based on the current data, the feedback cycle is short, which helps to provide more accurate decision basis for canteen management optimization, realizes the change from "lagging rectification" to "active prediction", quickly responds to the needs of diners, and continuously improves the canteen service quality and operation efficiency.

[0021] According to one aspect of the embodiments of the present application, a canteen satisfaction quantification evaluation method is provided. Figure 1A canteen satisfaction quantitative evaluation method according to one embodiment of the present application is shown. As shown in Figure 1 the method can include the following steps S110, S120, S130, S140, S150, S160 and S170.

[0022] In step S110, the Internet of Things sensing data and sales volume of each dish are acquired, and the Internet of Things sensing data includes the initial weight and the remaining amount of the corresponding dish.

[0023] In the scheme of the present example, the Internet of Things sensing data can be acquired by using Internet of Things devices such as smart plates with weighing function. That is, the single dish of a single diner is continuously tracked throughout the meal, and the weight change of each plate during the entire meal can also be counted, so as to calculate the actual remaining amount of each dish in the time interval and the overall remaining amount of the day. In some embodiments, the final weight of each plate can be automatically collected through Wi-Fi / Bluetooth protocol, the actual remaining amount of each dish / diner is calculated, and it is accurately associated to the corresponding dish ID. A standard data message protocol is defined, including device ID, timestamp, weight data, associated dish ID and other fields, to realize real-time and stable transmission of data. It can be understood that for a single dish, its sales volume, initial weight and remaining amount can be used as hard indicators of dish acceptance.

[0024] The dish sales volume can be automatically acquired from the canteen POS system. In some embodiments, it can be deeply interfaced with the existing consumption settlement system (POS machine) of the canteen. Each transaction stream data is automatically pulled at a fixed time or received in real time. Through database middleware (such as ODBC / JDBC) or calling the API interface provided by the POS manufacturer, the key data fields: transaction time, window ID, dish ID, dish name, sales quantity, consumption amount, payment method, etc. are acquired, which provide core data for sales analysis and preference analysis.

[0025] In step S120, based on the monitoring video stream of the canteen, the dining scene data is acquired, including queue detection data, regional personnel density data and seat occupancy rate of the canteen.

[0026] In the present example, the existing monitoring camera of the canteen can be used to perform real-time passenger flow statistics, queue length monitoring and personnel density analysis through computer vision algorithm, to quantify the smoothness of the dining experience. Specifically, the video stream of the canteen monitoring camera can be accessed, and the embedded or interfaced computer vision algorithm model is used to analyze the video stream in real time. For example, the existing passenger flow statistics and queue detection algorithm can be used to monitor the passenger flow and the queuing situation of each window in real time, to assist the operation decision. At the same time, the seat occupancy analysis algorithm can be used to monitor the vacancy rate of the dining area in real time, to evaluate the seat pressure.

[0027] At step S130, the active feedback text of the diners is acquired.

[0028] In the present example, an online scoring channel can be opened, for example, diners can score dishes and make text comments by scanning codes, applets, etc. The system for implementing the present solution can record user ID (which can be anonymized), score, comment content, time, associated dishes, etc.

[0029] At step S140, feedback information is extracted from the active feedback text, including dish score, dish comment, and canteen comfort score and comfort comment.

[0030] In the present example, a large language model, natural language processing technology (NLP), etc. can be used to automatically obtain dish score, dish comment, and canteen comfort score and comfort comment from active feedback text, which will not be described in detail.

[0031] In some embodiments, pre-trained natural language processing models can be used to obtain dish comments and comfort comments, which can include positive comments, negative comments, and neutral comments. In one specific embodiment, a pre-trained natural language processing model can be used to determine the sentiment polarity of the text comments in each active feedback text: whether the comments are "positive", "negative", or "neutral", and convert them into numerical values. Then, extract the core words in the comments (such as "too salty", "enough", "slow service"), and classify and count them to form word clouds or topic labels. This model can be trained in the following way: First, on a large general corpus (such as Wikipedia, news articles, web text), let the model pre-train through a self-supervised learning task (for example, randomly cover part of the words and let it predict) to master the basic grammar, semantics and world knowledge, and form a powerful general text understanding base (such as BERT or RoBERTa model). Then, for this specific scenario of canteen comments, collect and clean a large amount of dining-related comment data, and manually label its sentiment polarity (positive, negative, neutral) and key entities (such as dishes, taste, service). Finally, on this carefully annotated dining dataset, fine-tune the pre-trained model through supervised learning to make its parameters adapt and focus on identifying the sentiment tendency and domain keywords in dining comments, thereby obtaining a highly specialized sentiment analysis and keyword extraction engine.

[0032] As mentioned above, the sentiment polarity of text comments can be converted into numerical values. For example, "1" can represent positive, "0" represents neutral, and "-1" represents negative. For instance, comment A: "Kung Pao Chicken is so delicious, I'll order it again next time!" → The model judges it as positive → converted to a numerical value of +1; Comment B: "There were so many people today." → The model judges it as neutral → converted to a numerical value of 0; Comment C: "The rice was as hard as pebbles, inedible." → The model judges it as negative → converted to a numerical value of -1. This facilitates subsequent processing.

[0033] In some embodiments, core words in comments can be extracted using "text clustering" or "rule-based / dictionary" methods, and then categorized and statistically analyzed to form word clouds or topic tags. It is understood that the initially extracted keywords are scattered, and "text clustering" or "rule-based / dictionary" methods can group them into pre-defined, business-meaning topic categories. In a specific embodiment, extracting core words from comments and categorizing and statistically analyzing them to form word clouds or topic tags may include the following steps: First, extract the original keywords. Specifically, a natural language processing model can be used to extract the most core nouns, verbs, and adjectives from the comments. For example, for the feedback text "The braised pork is a bit too salty, and there's too much fat, but the portion is generous," the extracted keywords are: too salty, fat, too much, portion, generous. Then, the feedback text is categorized into topic categories. Specifically, a "business topic classification dictionary" can be pre-established to map keywords to high-level topics. For example, consider the following categories: Taste: too salty, bland, flavorful, spicy, delicious, greasy, strong flavor, tasteless; Portion: generous, small, large, affordable, not enough, too much; Ingredients: fresh, not fresh, too fatty, too lean, contains foreign objects, tough meat; Quality: delicious, unpalatable, authentic, genuine, so-so, good color, aroma, and taste; Service: slow service, good attitude, enthusiastic, indifferent, long queue; Environment: clean, tidy, crowded, noisy, has an odor, cool; Price: too expensive, cheap, affordable, high cost-performance ratio, not worth it. Once this data accumulates, the backend can generate valuable insights. For example, in negative reviews of "Kung Pao Chicken," 65% of the keywords were categorized under "#Taste#" as "too salty." In positive reviews of "All Dishes," 40% of the keywords were categorized under "#Portion#" as "generous portion." This information helps managers understand specific situations.

[0034] In step S150, based on the data collection time, dish name and window information, the IoT sensor data, sales volume, dining scenario data and feedback information are fused to obtain a multi-source data association model, which includes multiple sets of associated data.

[0035] It can be understood that different data collected in steps S110-S140 correspond to different time periods / dishes. In the present example, a data model with "time-dish-window" as the key dimension can be established, and all data are associated through key fields such as time stamp (which can be accurate to the minute), dish ID, window ID, etc. For example, "12:00 A dish sales" is associated with "12:00-12:15 queue length in this window" and "sentiment of comments about A dish on the same day" to form a set of associated data.

[0036] In one specific embodiment, the original data before fusion includes: Internet of Things sensor data (isolated and without business meaning), obtained at "2023-10-27 12:05:01", weight 45.2; video image recognition data (only reflecting macro state), obtained at "2023-10-27 12:30:00", queue length, 15, estimated waiting time 450; after ETL and data modeling center processing, the system forms one or more theme-oriented wide tables or data models. The following is an example of fusion data for the "dish analysis" theme: dish daily performance fact table: fusion sales indicators, cost and profit indicators, popularity indicators, waste and acceptance indicators, etc. Through data fusion, the system changes the weighing data from cold numbers to "hard indicators of dish acceptance". Video data changes from abstract queue length to "quantitative indicators of dining experience". Transaction data changes from simple sales records to "the basis of profitability analysis". Text comments change from subjective opinions to "quantifiable quality problem labels" through NLP. Finally, all these data are unified into a data model with dishes, time, windows, etc. as the core, forming clean and powerful data assets that can support intelligent decision-making.

[0037] In some embodiments, before data fusion, different data can be cleaned and standardized. Specifically, data from different sources and in different formats can be cleaned, denoised, formatted, and standardized to prepare for fusion analysis. Of course, different data can also be preprocessed to ensure data quality, such as de-duplication, filling missing values, correcting error values, formatting (such as uniform time format), etc.

[0038] In step S160, based on the multiple sets of associated data, the dish taste index of each dish, the service efficiency index of each window, and the environmental comfort of the canteen are determined.

[0039] After the data is fused, the dish taste index of each dish, the service efficiency index of each window, and the environment comfort of the canteen can be determined according to the respective associated data. For example, the dish taste index of each dish can be quantified based on dish ratings, sales, remaining meal quantities, etc., the service efficiency of each window can be determined based on the queue length, average queuing time, etc. in the queue detection data corresponding to each window, and the environment comfort can be determined based on some or all of the comfort ratings, comfort evaluations, regional personnel density data, and seat occupancy rates, etc. of the canteen. The specific determination method can be selected as needed, for example, the final result can be obtained by weighted sum, arithmetic sum, etc., which will not be described here.

[0040] In step S170, the weighted sum of the dish taste index, the service efficiency index, and the environment comfort is calculated to obtain the comprehensive satisfaction. The weight values of the dish taste index, the service efficiency index, and the environment comfort can be set according to the needs of the user (for example, the canteen manager), of course, they can also be dynamically adjusted according to the actual situation, which is not limited herein.

[0041] In the technical solution, the satisfaction degree of the canteen is evaluated based on the Internet of Things sensing data, sales and monitoring video stream of each dish and active feedback text of diners, wherein the multi-source passive data can reflect the real behavior of a large number of diners, and the active feedback text can reflect the subjective feelings, core demands and potential pain points of diners. On the one hand, the data from different sources and of different types are associated and integrated, which is beneficial to breaking the island and forming a unified satisfaction analysis view. On the other hand, the combination of the two can more objectively and comprehensively evaluate the satisfaction degree of the canteen, which can not only avoid subjective bias caused by the one-sided words of a small number of people, improve the scientificity and accuracy of the results, and on the other hand, the scheme can automatically generate the satisfaction degree of the canteen based on multi-source data, greatly reducing the manual intervention. Greatly improve the management efficiency and reduce the labor cost. And this method can dynamically monitor based on real-time data or quasi-real-time data stream, so as to output the comprehensive satisfaction degree in real time based on the current data, the feedback cycle is short, which helps to provide more accurate decision basis for canteen management optimization, realizes the fundamental change from "experience-driven" to "data-driven", "lagging rectification" to "active prediction", "fuzzy perception" to "accurate quantification", completely changes the dilemma of traditional canteen satisfaction management relying on subjective, lagging and one-sided information, and quickly responds to the needs of diners. In addition, this way builds a multi-dimensional quantitative index system covering dish quality (represented by dish taste index), service quality (represented by service efficiency index), and environment experience (represented by environment comfort), each index is supported by corresponding data source. This multi-dimensional and refined evaluation method can further improve the scientificity and accuracy of the results, and provide strong core support for the fine and intelligent operation of the canteen. It is especially suitable for scenes such as courts with fixed personnel structure, peak concentration and high privacy requirements. It can provide a more objective, real-time, comprehensive and in-depth satisfaction quantitative management and decision support tool on the basis of meeting the management specifications and security disciplines of government agencies, effectively improving the dining experience of diners and the management effect of the canteen, optimizing the operation cost and management efficiency of the canteen, and perfectly meeting the urgent needs of intelligent upgrading of the logistics support system in the construction of smart courts.

[0042] Illustratively, the dish taste index of each dish is determined, including: for any dish, based on the initial weight and the remaining amount of the dish, determining the remaining meal rate of the dish; based on the dish score of each diner for the dish, determining the average score of the dish; and according to the remaining meal rate, the average score and the sales of the dish, determining the dish taste index of the dish.

[0043] In the present example, for any dish, the leftover rate of each diner who purchased the dish can be calculated according to the initial weight of the dish and the leftover amount of the diner, and then the average of the leftover rates is calculated to obtain the leftover rate of the dish (which can be referred to as the normalized leftover rate for the sake of distinction). Similarly, the average score of the dish can be obtained by calculating the average of the scores of the dish, which is not described in detail.

[0044] After obtaining the normalized leftover rate, the average score and the sales of a specific dish, the dish taste index of the dish can be obtained by means of arithmetic sum or weighted sum. In one embodiment, the dish taste index can be obtained by the following formula: dish taste index = (sales weight * normalized sales) + (score weight * average score) - (leftover rate weight * normalized leftover rate).

[0045] In some embodiments, the leftover rate, the average score and the sales can also be normalized into normalized scores (e.g. 1-10 points) respectively, and then the weighted sum of the three is calculated to obtain the dish taste index. For example, for the leftover rate, a correspondence between different scores and different leftover rate ranges can be set in advance, and then the leftover score is determined according to the current normalized leftover rate. For the average score, it can be directly mapped to the range of [1, 10]. For the sales, the ratio of the sales of the dish to the total sales of the dish in the current dining time period can be calculated, and a mapping relationship between different ratio ranges and scores is established, and then the sales score is obtained according to the calculated ratio and the mapping relationship.

[0046] The above scheme calculates the dish taste index by combining the leftover rate of the dish (objectively reflecting the objective acceptance), the average score (reflecting the subjective satisfaction) and the sales (reflecting the market acceptance), realizes the quantitative and integrated evaluation of the taste and popularity of a single dish, breaks the one-sidedness of the traditional single index (such as only looking at sales or score), and provides a scientific basis for accurately judging the quality of the dish and optimizing the dish structure.

[0047] Exemplarily, the method further comprises: for each dish, determining whether the leftover rate of the dish is greater than a first leftover rate threshold; determining whether the dish taste index of the dish is less than a taste index threshold; when the leftover rate of the dish is greater than the first leftover rate threshold and the dish taste index of the dish is less than the taste index threshold, obtaining negative dish evaluations about the dish in the feedback information; and merging the leftover rate, the dish taste index and the negative dish evaluations of the dish into a transaction item set and outputting.

[0048] Optionally, the remaining meal rate, dish taste index and negative dish evaluation of the dish are combined into a transaction item set and output, which can specifically include the following steps: using association rule analysis (such as Apriori or FP-Growth algorithm) to mine the hidden relationship between discrete data items: in data processing, numerical data (such as remaining meal rate, score) is discretized into classification labels (such as “high remaining meal rate”, “taste score > 4.5”), and the keyword labels obtained by text analysis (such as “old ribs”, “slow service”) are combined to form a “transaction item set”. In some embodiments, after obtaining the transaction item set, the model can be used to further calculate the support, confidence and lift to output strong association rules, for example, it is found that {‘old ribs’, ‘salty’}=>{‘high remaining meal rate’} is a rule with high confidence. At this time, the confidence can be output together with the transaction item set. The transaction item set can provide a reliable basis for improving the kitchen of the canteen.

[0049] The above scheme accurately screens out problem dishes with “objective high remaining meal rate and poor comprehensive evaluation” by setting double thresholds of remaining meal rate and dish taste index, and forms a transaction item set combined with targeted negative evaluation, realizes the closed loop from “finding problem dishes” to “locating problem causes”, avoids the blindness of manual investigation, and significantly improves the efficiency and accuracy of dish problem diagnosis.

[0050] Illustratively, the method further includes predicting the demand quantity of the corresponding dish in the next meal period based on the remaining meal rate and the sales quantity of each dish. In some schemes, a time series prediction model can be constructed based on the historical data (including remaining meal rate and sales quantity) of each dish to predict future dish sales, peak flow, and even predict the potential popularity of new dishes after listing, providing decision support for precise meal preparation and dish iteration. In some embodiments, weather, workday type, etc. can also be combined for prediction to further improve prediction accuracy.

[0051] In this paper, the sales quantity of a dish can be the daily sales quantity or the average sales quantity in a period of time (one week, one month, etc.). Compared with the daily sales quantity, the average sales quantity in a period of time can better reflect the actual sales quantity, and therefore the average sales quantity is preferred.

[0052] Time series forecasting models include but are not limited to SARIMAX (seasonal autoregressive integrated moving average exogenous model), Facebook Prophet, machine learning models (such as LightGBM, XGBoost), etc. SARIMAX is a powerful tool for traditional time series analysis. It can explicitly handle seasonality (S), such as weekly repeating sales patterns. The X part allows us to add external variables, such as weather, holidays, etc. Facebook Prophet is a more modern, user-friendly time series forecasting model. It has built-in handling of seasonality and holiday effects, and is very robust to missing data and trend changes, making it very suitable for business forecasting scenarios. Machine learning models (such as LightGBM, XGBoost) can convert these time series problems into supervised learning problems. By constructing a feature dataset containing lag features (such as the past 7 days of sales), rolling statistical features (such as the average sales of the last 3 days), and all external variables, let the model learn its complex relationships by itself.

[0053] Of course, the above models can also be combined, for example, use Prophet or SARIMAX to capture the basic timing rules, and then use LightGBM to correct the residual and external variables.

[0054] In one specific embodiment, the canteen is a court canteen. In this embodiment, historical sales (the daily sales of each dish in the past weeks or months, which is the basis of the model.) can be used as the core time series data, workday type (Monday to Friday, weekend, pre-holiday, statutory holiday) as time feature data, semester cycle (for court canteens, exam week, training week, etc. are major influencing factors), environmental features (temperature, precipitation, weather phenomena, etc. weather data), dish information (dish price, dish type (such as noodles, rice), whether it is a new dish), predicted total number of diners, dish reputation index of the dish in the previous dining period (from sentiment analysis engine, such as "the average sentiment score of Kung Pao Chicken in the near future". A dish with a surge in negative reviews is likely to see a decline in sales the next day), historical leftover rate (average leftover rate of the dish in the past week), etc. as model input. Of course, considering the fixed nature of court dining personnel, we can focus on the dining needs of different groups (judges, police officers, other administrative personnel, etc.) according to type, thereby helping to create more diverse, rich, and accurate meals.

[0055] The above scheme predicts the demand for dishes in the next dining period based on historical leftover rates (reflecting the amount of leftovers) and sales (reflecting actual consumption), which can dynamically match supply and demand, reducing food waste caused by excessive preparation, avoiding insufficient supply affecting dining experience, achieving fine and intelligent canteen preparation, and reducing operating costs.

[0056] Exemplarily, the method further comprises: for each dish, generating a dish slow-selling warning when the remaining meal rate of the dish is greater than a second remaining meal rate threshold, and the sales volume of the dish is less than the historical average sales volume of the dish.

[0057] In this scheme, a warning rule can be set in advance, for example, "IF remaining meal rate > 30% AND sales volume < 50% of average sales volume THEN send 'dish slow-selling warning' to the administrator's mobile phone". In this way, a warning mechanism can be established based on real-time data flow to alarm abnormal situations (such as a sudden increase in the remaining meal rate) at a minute level, realizing active management. The administrator can respond immediately and intervene in processing before the problem expands (such as timely replenishing popular dishes), minimizing negative impact and improving operational efficiency.

[0058] Exemplarily, determining the service efficiency index of each window comprises: for any window, determining the average queuing time of the window based on the queue detection data corresponding to the window; and determining the service efficiency index based on the average queuing time.

[0059] In some embodiments, the service efficiency index can be calculated by the following formula: service efficiency index = (queue length weight * normalized queuing time).

[0060] The above scheme takes the window average queuing time as the core to quantify the service efficiency index, converts the abstract "service speed" into specific values that can be monitored and compared, and intuitively reflects the service level differences of each window, providing clear data support for optimizing window personnel allocation and shortening dining waiting time, and improving the objectivity and operability of service efficiency evaluation.

[0061] Exemplarily, the method further comprises: determining whether the average queuing time of each window exceeds a time threshold; and outputting a time warning when the average queuing time of any window exceeds the time threshold. In this way, the administrator can timely discover problems such as too many people queuing at the window, facilitating active intervention and rapid response (such as increasing the number of windows), especially suitable for monitoring and managing the service quality under extreme concentrated dining pressure.

[0062] Exemplarily, determining the environmental comfort of the canteen comprises: determining an average comfort score based on the comfort scores of each diner; counting the number of positive feedback keywords and negative feedback keywords in the comfort evaluation; determining a positive feedback ratio and a negative feedback ratio based on the number of positive feedback keywords and negative feedback keywords; and determining the environmental comfort based on the regional personnel density data, the seat occupancy rate, the average comfort score, the positive feedback ratio, and the negative feedback ratio (for example, an arithmetic sum or a weighted sum can be calculated).

[0063] The above scheme fuses the average comfort score (subjective feedback), the ratio of positive and negative keywords (text emotion), the regional personnel density and seat occupancy (objective environmental data) to calculate the environmental comfort, realizes the multi-dimensional comprehensive evaluation of the "subjective feeling + objective state" of the dining room environment, avoids the limitation of single data (such as only looking at the number of seats), and provides a comprehensive basis for improving the dining environment and improving the comfort.

[0064] Exemplarily, the method further comprises: fitting the service efficiency index obtained each time as the dependent variable, and the corresponding average queuing time, seat occupancy and number of work windows as the independent variables to obtain a regression equation; and performing regression analysis based on the regression equation to determine the influence degree of each independent variable on the service efficiency index.

[0065] It can be understood that for each independent variable, the coefficient of the independent variable can quantify the influence size and direction of the independent variable on the dependent variable. The above scheme quantifies the specific influence degree of each factor on the service efficiency (such as "increasing 1 window can make the efficiency index increase X%") by constructing the regression equation of the service efficiency index and the average queuing time, the seat occupancy and the number of work windows, breaks through the fuzziness of qualitative analysis, and provides precise quantitative guidance for scientifically formulating window scheduling, optimizing space layout and other operation strategies, and improves the scientificity and effectiveness of service optimization.

[0066] In some embodiments, the transaction item set scheme and the regression equation scheme described above can be applied simultaneously. In this way, panoramic problem diagnosis can be realized. Not only can it be known that "satisfaction has decreased", but also it can be accurately known whether it is "because the queue is too long" (service problem) or "because the ribs are overcooked" (quality problem), so that the deep reason for the decline in sales (such as: correlation analysis finds that the negative keyword "meat is not fresh") can be automatically diagnosed, and future trends can be predicted to support preventive management (such as: prepare meals in advance).

[0067] Exemplarily, the dining room is a court dining room, and the diners include judges, police officers and other administrative personnel of different groups. The method can further comprise: according to the group types of different diners, statistics the preference differences of different groups.

[0068] In some embodiments, different data can be collected through anonymization, aggregation and other technologies to ensure compliance with the confidentiality requirements of government agencies.

[0069] Exemplarily, the method further comprises: displaying the comprehensive satisfaction. In this embodiment, the core indexes such as the dish taste index, the service efficiency index, and the environment comfort level can also be displayed simultaneously, and a trend change graph, a data distribution graph, and a hotspot map constructed according to the canteen passenger flow can also be constructed according to historical data, which are not described herein. In a specific embodiment, the display device comprises but is not limited to a manager data cockpit (a visual large screen), a mobile terminal APP, and the like. The system homepage of the manager data cockpit can display the most key core indexes and real-time dynamics in a chart or other visual elements, including: real-time dining people, a current queue heat map, today's popular / unsalable dish TOP5, a satisfaction index trend graph, a pre-warning information list, and the like. The mobile terminal APP can push key pre-warning information; support viewing core charts on a mobile phone to realize "mobile office". Meanwhile, the system for implementing the method can also automatically generate detailed analysis reports at regular intervals to replace manual writing. Support generating daily reports, weekly reports, and monthly reports, the content including: data summary, trend analysis, problem diagnosis, improvement suggestions, and the like, and supporting one-key exporting PDF or automatically sending through email. In addition, the system can also be configured with user permission management (assigning viewing and operating permissions of different accounts), basic data maintenance (dish information, window information), pre-warning rule configuration, model parameter adjustment, and the like, which are not described herein.

[0070] According to still another aspect of the embodiments of the present application, an electronic device is also provided. Figure 2 A schematic block diagram of an electronic device according to an embodiment of the present application is shown. As shown, Figure 2 The electronic device 200 comprises a processor 210 and a memory 220. The memory 220 stores a computer program, and the processor 210 is configured to execute the computer program to implement the method described above.

[0071] According to still another aspect of the embodiments of the present application, a computer readable storage medium is also provided. The storage medium stores a computer program / instruction, which is executed by a processor to implement the method described above. The storage medium may, for example, include a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer readable storage medium can be any combination of one or more computer readable storage media.

[0072] Those skilled in the art can easily understand the implementation structure, working principle and beneficial effects of the electronic device and the computer readable storage medium by reading the above method. For brevity, they are not described herein.

[0073] Although example embodiments have been described herein with reference to the accompanying drawings, it is to be understood that the example embodiments are only exemplary and are not intended to limit the scope of the present disclosure. Those of ordinary skill in the art can make various changes and modifications of the example embodiments without departing from the scope and spirit of the present disclosure. All such changes and modifications are intended to be within the scope of the present disclosure as claimed.

[0074] Those skilled in the art can realize the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be realized by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the particular application and design constraints. Those skilled in the art can realize the described functions by various methods, and the present disclosure is not limited to a particular method for realizing the functions.

[0075] In several embodiments provided by the present disclosure, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, another division manner can be used, for example, a plurality of units or components can be combined or integrated into another device, or some features can be omitted or not performed.

[0076] In the specification provided herein, a large number of specific details are described. However, it can be understood that the embodiments of the present disclosure can be practiced without these specific details. In some examples, well-known methods, structures, and techniques are not described in detail in order not to obscure the understanding of the specification.

[0077] Similarly, it should be appreciated that, in the description of the example embodiments of the present disclosure, various features of the present disclosure are sometimes grouped together in a single embodiment, figure, or description of a related group of embodiments, for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various aspects, features and embodiments of the present disclosure. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed disclosure requires more features than are explicitly recited in each claim. Rather, inventive aspects lie in less than all features of any single disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim acting as a separate embodiment of the claimed disclosure. The claims are not to be interpreted as being limited to the embodiments or examples described herein.

[0078] Those skilled in the art will appreciate that all features described herein (including all features and processes described in the accompanying claims, abstract and drawings) can be combined in any combination. Each feature disclosed in this specification (including any "means for" feature disclosed by an "apparatus comprising a means for" claim), in the claims, abstract and drawings can be replaced by alternative features that are both equivalent in terms of the functionality for which the features are described to perform. This applies no matter whether the alternative is prior art to, or prior to, the disclosed embodiment.

[0079] Furthermore, those skilled in the art will appreciate that the features described herein, although characterized as being included in some embodiments and not others, are combinable in different embodiments to form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0080] Embodiments of various components of the present application can be implemented in hardware, software, or a combination thereof. Those skilled in the art will appreciate that some or all of the functionality of some of the modules in the electronic devices according to embodiments of the present application can be implemented using a microprocessor or a digital signal processor (DSP) in practice. The present application can also be implemented as a program (for example, a computer program and a computer program product) for executing some or all of the methods described herein. Such a program implementing the present application can be stored on a computer readable medium or can have one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or in any other form.

[0081] It is noted that the foregoing examples have been provided merely for the purpose of explanation and are in no way to be construed as limiting of the present application. While the present application has been described with reference to specific embodiments, it is apparent that various modifications and changes can be made by those skilled in the art without departing from the scope of the appended claims. In its broadest form, the present application is defined by the appended claims and equivalents thereof and all of the additional features which fall within the scope of the claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The application can be implemented by means of both hardware and software, and any combination thereof. In the claims, the word "first", "second", "third", etc. does not imply any order. The terms "first", "second", "third", etc. should be interpreted as names.

[0082] The above merely describes specific embodiments or specific implementation of the present application, and the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. The protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for quantitatively evaluating canteen satisfaction, characterized by, The method comprises: acquiring Internet of Things sensor data and sales of each dish, the Internet of Things sensor data including initial weight and remaining amount of the corresponding dish; acquiring dining scene data based on the monitoring video stream of the canteen, the dining scene data including queue detection data, regional personnel density data and seat occupancy rate of the canteen; acquiring active feedback text of diners; extracting feedback information from the active feedback text, the feedback information including dish rating, dish evaluation and comfort rating and comfort evaluation of the canteen; performing data fusion on the Internet of Things sensor data, the sales, the dining scene data and the feedback information according to data collection time, dish name and window information to obtain a multi-source data association model, the multi-source data association model including multiple sets of associated data; determining dish taste index of each dish, service efficiency index of each window and environmental comfort of the canteen based on the multiple sets of associated data; calculating weighted sum of the dish taste index, the service efficiency index and the environmental comfort to obtain comprehensive satisfaction.

2. The method of claim 1, wherein, The method further comprises: for any dish, determining remaining meal rate of the dish based on initial weight and remaining amount of the dish; determining average rating of the dish based on dish rating of each diner to the dish; determining dish taste index of the dish according to the remaining meal rate, the average rating and sales of the dish.

3. The method of claim 2, wherein, The method further comprises: for each dish, judging whether the remaining meal rate of the dish is greater than a first remaining meal rate threshold; judging whether the dish taste index of the dish is less than a taste index threshold; when the remaining meal rate of the dish is greater than the first remaining meal rate threshold and the dish taste index of the dish is less than the taste index threshold, acquiring negative dish evaluation about the dish in the feedback information; merging the remaining meal rate, the dish taste index and the negative dish evaluation of the dish into a transaction item set and outputting.

4. The method of claim 2, wherein, The method further comprises: predicting demand of the corresponding dish in the next meal period based on the remaining meal rate and the sales of each dish.

5. The method of claim 2, wherein, The method further comprises: for each dish, generating dish slow-selling warning when the remaining meal rate of the dish is greater than a second remaining meal rate threshold and the sales of the dish is less than the historical average sales of the dish.

6. The method according to any one of claims 1 to 5, characterized in that, The method further comprises: for any window, determining average queuing time of the window based on queue detection data corresponding to the window; determining the service efficiency index based on the average queuing time.

7. The method according to any one of claims 1 to 5, characterized in that, The method further comprises: determining average comfort rating based on comfort rating of each diner; counting number of positive feedback keywords and negative feedback keywords in the comfort evaluation; determining positive feedback ratio and negative feedback ratio based on the number of positive feedback keywords and the number of negative feedback keywords; determining the environmental comfort based on the regional personnel density data, the seat occupancy rate, the average comfort rating, the positive feedback ratio and the negative feedback ratio.

8. The method according to any one of claims 1 to 5, characterized in that, The method further comprises: The service efficiency index obtained each time is taken as a dependent variable, and the corresponding average queuing time, seat occupancy rate and number of working windows are taken as independent variables to fit to obtain a regression equation; Regression analysis is performed based on the regression equation to determine the influence degree of each independent variable on the service efficiency index.

9. An electronic device, comprising: A processor and a memory are included, the memory stores a computer program, and the processor is configured to execute the computer program to implement the method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, A computer program / instruction is stored, and the computer program / instruction is executed by a processor to implement the method according to any one of claims 1-8.

Images