Marketing estimation software and estimation algorithm thereof

Abstract

The invention discloses marketing estimation software and an estimation algorithm thereof. The estimation algorithm comprises the use of a prediction tool, the setting of the prediction tool and the setting of a data worksheet and a date worksheet. The method further comprises a time sequence prediction algorithm and a feature extraction algorithm. According to the method, a time sequence prediction algorithm and a feature extraction algorithm are used, a given budget is firstly subjected to budget arrangement according to historical cancel-after-verification condition feature conditions, then the budget is used as a variable factor to construct the time sequence prediction algorithm, trend and period features of a whole sales model are fused, and then full-amount GMV prediction of a next period is achieved. And the error is controlled within 10%.

Description

Technical Field

[0001] The present invention relates to the technical field of marketing estimation, and in particular to marketing estimation software and an estimation algorithm thereof. Background Art

[0002] Marketing budget refers to the optimal amount of budget required to implement various marketing strategies and policies, as well as the budget allocation between various marketing links and various marketing methods. Marketing budget usually consists of three parts: sales revenue budget, sales cost budget, and marketing expense budget. In addition to these three parts, the company's operating budget also includes administrative expense budget, research and development expense budget, tax budget and other indicators.

[0003] In traditional marketing methods, merchants estimate the total merchandise transaction amount that the invested budget will generate based on experience. This will result in a large error in the total merchandise transaction amount and may also lead to insufficient investment, thus affecting normal marketing. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a marketing estimation software and an estimation algorithm thereof.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a marketing estimation software, including the use of forecasting tools, the setting of forecasting tools and the setting of data worksheets and date worksheets

[0006] The prediction tool is used by opening the prediction tool software on a computer, and the prediction tool interface will pop up after the command prompt is loaded;

[0007] The setting of the forecasting tool includes selecting an Excel table for storing data at the data file, selecting the data to be used in the data tool table options, selecting an Excel table for storing dates at the date file, selecting the date data to be used in the date tool table options, filling in the corresponding forecast time for the forecast year and month, setting the options and clicking Run to set the location to save the Excel table, a completion prompt box will pop up after the calculation is completed, setting the forecast for the next month or future budget, outputting the results after the operation is completed, setting the accuracy of the verification algorithm, and the data file contains the data for the forecast year and month, and outputting the results after the operation is completed;

[0008] The settings of the data worksheet and the date worksheet include data worksheet header settings and date header settings. The data worksheet header settings include date, total GMV, active GMV, write-off and series, and the date header settings include date and budget calendar.

[0009] A method for using marketing estimation software comprises the following steps:

[0010] 1) Select the historical data file through "Data File";

[0011] 2) Select the corresponding Excel worksheet in the data file;

[0012] 3) Select the calendar file through "Date File";

[0013] 4) After selecting the file, select the corresponding Excel worksheet;

[0014] 5) Select the data range used for training the model. Select "Up to the forecast month" to use only the data before the forecast month for training. Select "All" to use all the provided data for training.

[0015] 6) Enter the year and month for the required forecast;

[0016] 7) Select the budget method. The data file contains the date data of the forecast year and month. If you want to get the forecast month budget directly through statistical historical data, select "Historical Data Import". If you need to set a custom budget value, select "Manual Input";

[0017] 8) When selecting "Manual Input", you need to fill in the budget amount in "Budget";

[0018] 9) Select the output method. If the data file contains the date data of the forecast year and month, if you want to compare the forecast results with historical data, select "Compare with historical data". If you select "Forecast results only", only the forecast results will be output without comparing with historical data.

[0019] 10) Click "Run", select the location to save the output results, and wait for the calculation to complete.

[0020] An estimation algorithm for marketing estimation software, including a time series prediction algorithm and a feature extraction algorithm,

[0021] The feature extraction algorithm refers to budgeting a given budget according to the historical write-off characteristics. The obtained budget characteristics will provide external influencing factor information for subsequent algorithms.

[0022] The time series prediction algorithm refers to using the obtained data to arrange them in sequence in chronological order, analyzing their changing direction and degree, and thus speculating on the levels that may be achieved in several future periods; using the autoregressive integrated moving average model with seasonality and additional inputs as the basic model.

[0023] As a further description of the above technical solution:

[0024] The feature extraction algorithm includes the following steps:

[0025] Calculate the average write-off ratio of each budget time point type in the same period of the previous year, that is, the same month of the previous year, the previous month of the previous year, and the last month of the previous year. The budget type set for the same period of the previous year is recorded as X.

[0026] Obtain the write-off data for each budget time period in the same period of the previous year, and obtain the total write-off amount V for each budget type time point according to the budget calendar. i ;

[0027] Obtain the write-off data for each budget time period in the same period of the previous year, and calculate the duration T of each budget type time point according to the budget calendar. i ;

[0028] The average write-off value V' from the statistical time to each budget time period i formula:

[0029]

[0030] Taking working days as the benchmark value, calculate the ratio of each budget time period to working days P i :

[0031]

[0032] Calculate the budget ratio of the month that needs to be predicted, and record the budget type set of the required and predicted months as Y;

[0033] According to the budget calendar of the forecast month, determine the specific number of days N in each budget time period j ;

[0034] Multiply the previous year's write-off ratio by the number of days in the current cycle to calculate the budget ratio R for each budget calendar in the current cycle. j ;

[0035] R j =P j ×N i (i∈X,j∈Y,i=j)

[0036] For budget types that did not appear in the same period of the previous year, Optionally, directly change R j =1.2;

[0037] Calculate the allocated budget for each budget calendar based on the ratio, and then divide it by the number of days to get the specific allocated budget for each day;

[0038] According to the above budget ratio, calculate the budget Q of each budget calendar on weekdays 工作日 ;

[0039]

[0040] Calculate the monthly and daily budgets that need to be forecasted d , let the budget type corresponding to day d be f(d);

[0041] B d =Q 工作日 *R f(d) .

[0042] As a further description of the above technical solution:

[0043] The time series prediction algorithm refers to the SARIMAX formula:

[0044]

[0045] The time series prediction algorithm comprises the following steps:

[0046] Set the preliminary parameters, cycle parameters, seasonal impact parameters, and external factor parameters:

[0047] p: autoregressive order;

[0048] d: difference order;

[0049] q: moving average order;

[0050] P: seasonal autoregressive order;

[0051] D: seasonal difference order;

[0052] Q: seasonal moving average order;

[0053] s: length of seasonal cycle;

[0054] Since the data in the same period last year and the past three months have a significant impact on the model changes, these data are used as test data, and the optimizer is used to train the model to find new and possibly better parameters:

[0055] The data of the same period last year and the past three months are used as the test data of the model;

[0056] The budget allocation for each month is calculated separately as an exogenous variable;

[0057] When the same set of parameters is used, the monthly data of the test data and the exogenous variables are used to make predictions using SARIMAX as the estimated results of the parameter;

[0058] The mean absolute error between the estimated results obtained by comparing different parameter combinations and the actual test data is used as an indicator to measure the excellence of the parameters;

[0059] Use the optimizer to train the model, adjust parameters, and find a possible better parameter combination;

[0060] Define the parameter search scope:

[0061] p range: [1,31];

[0062] d range: [0,2];

[0063] q range: [1,14];

[0064] P range: [0,2];

[0065] D range: [0,2];

[0066] Q range: [0,2];

[0067] Select the parameter combination with the smallest error and determine the model parameters to be used;

[0068] Use the parameters and data to get the prediction model, use the time series prediction algorithm SARIMAX to make predictions, and get the prediction results:

[0069] Using the finalized parameters and integrated data, a time series forecasting model is constructed;

[0070] Use the time series prediction algorithm SARIMAX to predict and get the prediction results of future GMV;

[0071] After obtaining the forecast results, further calculate whether there is a significant change trend in the current period's write-offs compared with the same period last year. If so, adjust the results according to the change ratio:

[0072] Compare the forecast results with the write-off data of the same period last year, analyze whether there is a significant change trend, and calculate the change rate:

[0073] If the rate of change ΔR exceeds the preset threshold, the prediction result is adjusted according to the part of the change ratio ΔR that exceeds the threshold. The adjustment formula is:

[0074] Prediction results 调整后 = predicted result × (1 + part exceeding the threshold).

[0075] The present invention has the following beneficial effects:

[0076] Compared with the prior art, the present invention uses a time series prediction algorithm and a feature extraction algorithm to first arrange a given budget according to the historical write-off characteristics, and then constructs a time series prediction algorithm by taking the budget as a variable factor, integrating the trend and cycle characteristics of the entire sales model, thereby realizing the full GMV prediction for the next cycle with an error controlled within 10%. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a flow chart of the time series prediction algorithm of the present invention;

[0078] Figure 2 It is a flow chart of the feature extraction algorithm of the present invention. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] A marketing estimation software, including the use of forecasting tools, the settings of forecasting tools and the settings of data worksheets and date worksheets

[0081] Use the prediction tool to open the prediction tool software on your computer. After the command prompt is loaded, the prediction tool interface will pop up.

[0082] The settings of the forecasting tool include selecting the Excel table for storing data in the data file, selecting the data to be used in the data tool table options, selecting the Excel table for storing dates in the date file, selecting the date data to be used in the date tool table options, filling in the corresponding forecast time for the forecast year and month, setting the options and clicking Run to set the location to save the Excel table, a completion prompt box will pop up after the calculation is completed, setting the forecast for the next month or future budget, outputting the results after the operation is completed, setting the accuracy of the verification algorithm, and the data file contains the data for the forecast year and month, and outputting the results after the operation is completed;

[0083] The settings of the data worksheet and date worksheet include the data worksheet header settings and the date header settings. The data worksheet header settings include date, total GMV, active GMV, write-off and series, and the date header settings include date and budget calendar.

[0084] A method for using marketing estimation software comprises the following steps:

[0085] 1) Select the historical data file through "Data File"; specifically, the data file needs to have at least one year of data; the data should contain continuous data up to the month before the forecast as much as possible to make a coherent forecast; the data should be as uninterrupted as possible and should contain data from the same period last year for budget allocation.

[0086] 2) Select the corresponding Excel worksheet in the data file;

[0087] 3) Select the calendar file through "Date File"; when using, the date and budget calendar columns cannot be empty.

[0088] 4) After selecting the file, select the corresponding Excel worksheet;

[0089] 5) Select the data range used for training the model. Select "Up to the forecast month" to use only the data before the forecast month for training. Select "All" to use all the provided data for training. When the total GMV, activity GMV, and write-off columns are empty, they will be calculated as 0. The calculation of data that has integrated all series is supported, that is, the series column does not need to exist.

[0090] 6) Enter the year and month for the required forecast;

[0091] 7) Select the budget method. The data file contains the date data of the forecast year and month. If you want to get the forecast month budget directly through statistical historical data, select "Historical Data Import". If you need to define your own budget value, select "Manual Input". When you select "Historical Data" in "Budget Method", the data file needs to include the date data of the forecast year and month.

[0092] 8) When selecting "Manual Input", you need to fill in the budget amount in "Budget";

[0093] 9) Select the output method. When the data file contains the date data of the forecast year and month, if you want to compare the forecast results with historical data, select "Compare with historical data". If "Forecast results only" is selected, only the forecast results will be output without comparing with historical data. Select "Compare with historical data" in "Output method". The data file needs to include the date data of the forecast year and month.

[0094] 10) Click "Run", select the location to save the output results, and wait for the calculation to complete.

[0095] An estimation algorithm for marketing estimation software, including a time series prediction algorithm and a feature extraction algorithm,

[0096] The feature extraction algorithm is to arrange the given budget according to the historical write-off characteristics. The obtained budget characteristics will provide external influencing factor information for subsequent algorithms.

[0097] The time series forecasting algorithm refers to the use of the obtained data to arrange them in chronological order, analyze their direction and degree of change, and thus speculate on the levels that may be achieved in several future periods; using the autoregressive integrated moving average model with seasonality and additional inputs as the basic model.

[0098] The feature extraction algorithm consists of the following steps,

[0099] Calculate the average write-off ratio of each budget time point type in the same period of the previous year, that is, the same month of the previous year, the previous month of the previous year, and the last month of the previous year. The budget type set for the same period of the previous year is recorded as X.

[0100] Obtain the write-off data for each budget time period in the same period of the previous year, and obtain the total write-off amount V for each budget type time point according to the budget calendar. i ;

[0101] Obtain the write-off data for each budget time period in the same period of the previous year, and calculate the duration T of each budget type time point according to the budget calendar. i ;

[0102] The average write-off value V' from the statistical time to each budget time period i formula:

[0103]

[0104] Taking working days as the benchmark value, calculate the ratio of each budget time period to working days P i :

[0105]

[0106] Calculate the budget ratio of the month that needs to be predicted, and record the budget type set of the required and predicted months as Y;

[0107] According to the budget calendar of the forecast month, determine the specific number of days N in each budget time period j ;

[0108] Multiply the previous year's write-off ratio by the number of days in the current cycle to calculate the budget ratio R for each budget calendar in the current cycle. j ;

[0109] R j =P j ×N i (i∈X,j∈Y,i=j)

[0110] For budget types that did not appear in the same period of the previous year, Optionally, directly change R j =1.2;

[0111] Calculate the allocated budget for each budget calendar based on the ratio, and then divide it by the number of days to get the specific allocated budget for each day;

[0112] According to the above budget ratio, calculate the budget Q of each budget calendar on weekdays 工作日 ;

[0113]

[0114] Calculate the monthly and daily budgets that need to be forecasted d , let the budget type corresponding to day d be f(d);

[0115] B d =Q 工作日 *R f(d) .

[0116] The time series prediction algorithm refers to the SARIMAX formula:

[0117]

[0118] The time series forecasting algorithm consists of the following steps,

[0119] Set the preliminary parameters, cycle parameters, seasonal impact parameters, and external factor parameters:

[0120] pAutoregression order: 7 (indicates that the forecast is made using the data from the past 7 days)

[0121] d Difference order: 2 (indicates the use of second-order differences to make the time series stationary)

[0122] q Moving average order: 7 (using the average of the past 7 days)

[0123] PSeasonal autoregressive order: 1

[0124] D Seasonal difference order: 1

[0125] Q seasonal moving average order: 1

[0126] s Length of seasonal cycle: 365 (annual cycle).

[0127] Since the data in the same period last year and the past three months have a significant impact on the model changes, these data are used as test data, and the optimizer is used to train the model to find new and possibly better parameters:

[0128] The data of the same period last year and the past three months are used as the test data of the model;

[0129] The budget allocation for each month is calculated separately as an exogenous variable;

[0130] When the same set of parameters is used, the monthly data of the test data and the exogenous variables are predicted using SARIMAX as the estimated results of the parameter;

[0131] The mean absolute error between the estimated results obtained by comparing different parameter combinations and the actual test data is used as an indicator to measure the excellence of the parameters;

[0132] Use the optimizer to train the model, adjust parameters, and find a possible better parameter combination;

[0133] Define the parameter search scope:

[0134] p range: [1,31];

[0135] d range: [0,2];

[0136] q range: [1,14];

[0137] P range: [0,2];

[0138] D range: [0,2];

[0139] Q range: [0,2];

[0140] Select the parameter combination with the smallest error and determine the model parameters to be used;

[0141] Use the parameters and data to get the prediction model, use the time series prediction algorithm SARIMAX to make predictions, and get the prediction results:

[0142] Using the finalized parameters and integrated data, a time series forecasting model is constructed;

[0143] Use the time series prediction algorithm SARIMAX to predict and get the prediction results of future GMV;

[0144] After obtaining the forecast results, further calculate whether there is a significant change trend in the current period's write-offs compared with the same period last year. If so, adjust the results according to the change ratio:

[0145] Compare the forecast results with the write-off data of the same period last year, analyze whether there is a significant change trend, and calculate the change rate:

[0146] If the rate of change ΔR exceeds the preset threshold, the prediction result is adjusted according to the part of the change ratio ΔR that exceeds the threshold. The adjustment formula is:

[0147] Prediction results 调整后 = predicted result × (1 + part exceeding the threshold).

[0148] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A marketing estimation software, characterized by: Including the use of forecasting tools, the settings of forecasting tools and the settings of data worksheets and date worksheets, The prediction tool is used by opening the prediction tool software on a computer, and the prediction tool interface will pop up after the command prompt is loaded; The setting of the forecasting tool includes selecting an Excel table for storing data at the data file, selecting the data to be used in the data tool table options, selecting an Excel table for storing dates at the date file, selecting the date data to be used in the date tool table options, filling in the corresponding forecast time for the forecast year and month, setting the options and clicking Run to set the location to save the Excel table, a completion prompt box will pop up after the calculation is completed, setting the forecast for the next month or future budget, outputting the results after the operation is completed, setting the accuracy of the verification algorithm, and the data file contains the data for the forecast year and month, and outputting the results after the operation is completed; The settings of the data worksheet and the date worksheet include data worksheet header settings and date header settings. The data worksheet header settings include date, total GMV, active GMV, write-off and series, and the date header settings include date and budget calendar.

2. A method for using marketing estimation software, characterized in that: The following steps are included: 1) Select the historical data file through "Data File"; 2) Select the corresponding Excel worksheet in the data file; 3) Select the calendar file through "Date File"; 4) After selecting the file, select the corresponding Excel worksheet; 5) Select the data range to be used for training the model. Select "Up to the forecast month" to use only the data before the forecast month for training. Select "All" to use all the provided data for training. 6) Enter the year and month for the required forecast; 7) Select the budget method. The data file contains the date data of the forecast year and month. If you want to get the forecast month budget directly by statistical historical data, select "Historical Data Import". If you need to set a custom budget value, select "Manual Input". 8) When selecting "Manual Input", you need to fill in the budget amount in "Budget"; 9) Select the output method. If the data file contains the date data of the forecast year and month, if you want to compare the forecast results with historical data, select "Compare with historical data". If you select "Forecast results only", only the forecast results will be output without comparing with historical data. 10) Click "Run", select the location to save the output results, and wait for the calculation to complete.

3. An estimation algorithm for marketing estimation software, characterized by: Including time series prediction algorithms and feature extraction algorithms, The feature extraction algorithm refers to budgeting a given budget according to the historical write-off characteristics. The obtained budget characteristics will provide external influencing factor information for subsequent algorithms. The time series prediction algorithm refers to using the obtained data to arrange them in sequence in chronological order, analyzing their changing direction and degree, and thus speculating on the levels that may be achieved in several future periods; using the autoregressive integrated moving average model with seasonality and additional inputs as the basic model.

4. The estimation algorithm of the marketing estimation software according to claim 3, characterized in that: The feature extraction algorithm includes the following steps: Calculate the average write-off ratio of each budget time point type in the same period of the previous year, that is, the same month of the previous year, the previous month of the previous year, and the last month of the previous year. The budget type set for the same period of the previous year is recorded as X. Obtain the write-off data for each budget time period in the same period of the previous year, and obtain the total write-off amount V for each budget type time point according to the budget calendar. i ; Obtain the write-off data for each budget time period in the same period of the previous year, and calculate the duration T of each budget type time point according to the budget calendar. i ; The average write-off value V' from the statistical time to each budget time period i formula: Taking working days as the benchmark value, calculate the ratio of each budget time period to working days P i : Calculate the budget ratio of the month that needs to be predicted, and record the budget type set of the required and predicted months as Y; According to the budget calendar of the forecast month, determine the specific number of days N in each budget time period j ; Multiply the previous year's write-off ratio by the number of days in the current cycle to calculate the budget ratio R for each budget calendar in the current cycle. j ; R j =P j ×N i (i∈X,j∈Y,i=j) For budget types that did not appear in the same period of the previous year, Optionally, directly change R j =1.2; Calculate the allocated budget for each budget calendar based on the ratio, and then divide it by the number of days to get the specific allocated budget for each day; According to the above budget ratio, calculate the budget Q of each budget calendar on weekdays 工作日 ; Calculate the monthly and daily budgets that need to be forecasted d , let the budget type corresponding to day d be f(d); B d =Q 工作日 *R f(d) 。 5. The estimation algorithm of the marketing estimation software according to claim 3, characterized in that: The time series prediction algorithm refers to the SARIMAX formula: The time series prediction algorithm comprises the following steps: Set the preliminary parameters, cycle parameters, seasonal impact parameters, and external factor parameters: p: autoregressive order; d: difference order; q: moving average order; P: seasonal autoregressive order; D: seasonal difference order; Q: seasonal moving average order; s: length of seasonal cycle; Since the data in the same period last year and the past three months have a significant impact on the model changes, these data are used as test data, and the optimizer is used to train the model to find new and possibly better parameters: The data of the same period last year and the past three months are used as the test data of the model; The budget allocation for each month is calculated separately as an exogenous variable; When the same set of parameters is used, the monthly data of the test data and the exogenous variables are predicted using SARIMAX as the estimated results of the parameter; The mean absolute error between the estimated results obtained by comparing different parameter combinations and the actual test data is used as an indicator to measure the excellence of the parameters; Use the optimizer to train the model, adjust parameters, and find a possible better parameter combination; Define the parameter search scope: p range: [1,31]; d range: [0,2]; q range: [1,14]; P range: [0,2]; D range: [0,2]; Q range: [0,2]; Select the parameter combination with the smallest error and determine the model parameters to be used; Use the parameters and data to get the prediction model, use the time series prediction algorithm SARIMAX to make predictions, and get the prediction results: Using the finalized parameters and integrated data, a time series forecasting model is constructed; Use the time series prediction algorithm SARIMAX to predict and get the prediction results of future GMV; After obtaining the forecast results, further calculate whether there is a significant change trend in the current period's write-offs compared with the same period last year. If so, adjust the results according to the change ratio: Compare the forecast results with the write-off data of the same period last year, analyze whether there is a significant change trend, and calculate the change rate: If the rate of change ΔR exceeds the preset threshold, the prediction result is adjusted according to the part of the change ratio ΔR that exceeds the threshold. The adjustment formula is: Prediction results 调整后 = predicted result × (1 + part exceeding the threshold).

Images