A procurement big data management system

Abstract

A procurement big data management system includes the following modules: a storage module, the storage module is used to store procurement information, the procurement information includes procurement instructions, and is also used to store the procurement content similarity between at least two procurement information, and a procurement similarity value calculation module is used to execute the steps of selecting the core content of a first procurement instruction to obtain a first content, performing text segmentation on the first content to obtain a first segmentation result, and constructing a first content word vector according to the first segmentation result, wherein the first content word vector includes each segmentation and the word frequency of each segmentation. The above technical solution can store multiple procurement information, calculate the procurement content similarity between each node, and quickly determine the correlation between multiple procurement processes.

Description

Technical Field

[0001] The present invention relates to the field of procurement data management, and in particular to a management system involving procurement big data. Background Art

[0002] Enterprise informatization construction improves the production and operation efficiency of enterprises, reduces operational risks and costs, and thus improves the overall management level and sustainable operation capabilities of enterprises through the deployment of computer technology. At present, many enterprises purchase various information systems through procurement. The quality of the suppliers of enterprise information systems is closely related to the success of the information project. The selection of procurement suppliers and the signing of procurement contracts largely determine the survival and sustainable development of enterprises. Therefore, how to evaluate suppliers has become a research topic.

[0003] At present, enterprises basically formulate corresponding qualification conditions for supplier evaluation, and then use experts to evaluate various materials provided by suppliers that meet the qualification conditions, including: enterprise qualifications, procurement system software, procurement system demonstration, etc. However, the management of records related to suppliers and procurement processes has not yet been formed. Summary of the invention

[0004] To this end, it is necessary to provide a solution that can systematically and scientifically manage procurement big data.

[0005] To achieve the above object, the inventor provides a procurement big data management system, comprising the following modules: a storage module, the storage module is used to store procurement information, the procurement information includes procurement instructions, and is also used to store the procurement content similarity between at least two procurement information,

[0006] A procurement similarity value calculation module, the procurement similarity value calculation module is used to execute the steps of selecting the core content of the first procurement description to obtain the first content, performing text segmentation on the first content to obtain the first segmentation result, constructing a first content word vector according to the first segmentation result, the first content word vector includes each segmentation and the word frequency of each segmentation, selecting the core content of the second procurement description to obtain the second content, performing text segmentation on the second content to obtain the second segmentation result, constructing a second core content word vector according to the second segmentation result, the second content word vector includes each segmentation and the word frequency of each segmentation,

[0007] The purchase similarity value calculation module is used to calculate the purchase content similarity according to the cosine similarity or simple common word similarity between the first content word vector and the second content word vector.

[0008] Specifically, the storage module is also used to store enterprise information, and also stores procurement information and the connection between participating enterprises.

[0009] Specifically, after receiving new procurement information, the storage module calculates the procurement content similarity between the new procurement information and the stored procurement information if the participating enterprise of the new procurement information is the same as at least one participating enterprise of the stored procurement information, and saves the procurement content similarity between the new procurement information and the stored procurement information.

[0010] Specifically, after receiving the new purchase information, the storage module calculates the purchase content similarity between the new purchase information and the stored purchase information, and saves the purchase content similarity between the new purchase information and the stored purchase information.

[0011] Furthermore, the procurement similarity value calculation module is used to select the core content of the procurement instructions, which specifically includes the steps of identifying the templated content in the procurement instructions, including the tender notice, tender instructions, contract terms, and tender document format requirements, deleting the templated content, and obtaining the selected content.

[0012] Specifically, part-of-speech tagging is performed, and the weights of the noun vectors and the verb vectors in the first content word vector and the second content word vector are increased.

[0013] Furthermore, it also includes a title recognition module, which is used to use regular expressions to perform title recognition on the first content and the second content, wherein the title includes a main title, a subtitle, a main title and a subtitle; perform word segmentation and part-of-speech tagging on the title,

[0014] The purchase similarity value calculation module is used to add weight to the word vector obtained by the title sentence segmentation in the first content word vector; and is also used to add weight to the title sentence segmentation vector in the second content word vector.

[0015] Furthermore, it also includes a supplier scoring module, which is used to calculate the supplier score of the enterprise relative to each procurement information in which it participates according to preset rules.

[0016] Specifically, the storage module is also used to store supplier ratings between the enterprise and the procurement information in which the enterprise participates.

[0017] Different from the existing technology, the above technical solution can store multiple procurement information, take each procurement information as a node, and use the procurement instructions in the procurement information as the basis. It can calculate the similarity of procurement content between each node, so as to quickly determine the correlation between multiple procurement processes and provide a reference for the overall evaluation of the procurement system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a module diagram of the procurement big data management system described in the specific implementation method;

[0019] Figure 2 This is a diagram of a method for calculating the similarity of procurement content as described in a specific implementation method.

[0020] Description of reference numerals:

[0021] 101. Storage module,

[0022] 102. Purchase similarity value calculation module;

[0023] 103. Supplier rating module. DETAILED DESCRIPTION

[0024] In order to explain the technical content, structural features, achieved objectives and effects of the technical solution in detail, the following is a detailed description in conjunction with specific embodiments and accompanying drawings.

[0025] Please see here Figure 1 , is a procurement big data management system, comprising the following modules, a storage module 100, the storage module is used to store procurement information, the procurement information includes procurement instructions, and is also used to store procurement content similarity between at least two procurement information,

[0026] A procurement similarity value calculation module 102, the procurement similarity value calculation module is used to execute the steps of selecting the core content of the first procurement description to obtain the first content, performing text segmentation on the first content to obtain the first segmentation result, constructing a first content word vector according to the first segmentation result, the first content word vector includes each segmentation and the word frequency of each segmentation, selecting the core content of the second procurement description to obtain the second content, performing text segmentation on the second content to obtain the second segmentation result, constructing a second core content word vector according to the second segmentation result, the second content word vector includes each segmentation and the word frequency of each segmentation, and the procurement similarity value calculation module is used to calculate the procurement content similarity according to the cosine similarity or simple common word similarity between the first content word vector and the second content word vector.

[0027] The above technical solution can store multiple procurement information, take each procurement information as a node, and take the procurement instructions in the procurement information as the basis. It can calculate the similarity of procurement content between each node, so as to quickly determine the correlation between multiple procurement processes and provide a reference for the overall evaluation of the procurement system.

[0028] See also Figure 2 The procurement content similarity evaluation method performed by the procurement similarity value calculation module 102 of this embodiment includes the following steps:

[0029] S101 selects the core content of the first procurement description to obtain the first content, performs text segmentation on the first content to obtain the first segmentation result, and constructs a first word vector based on the first segmentation result, wherein the first word vector includes each segmentation and the word frequency of each segmentation. S102 selects the core content of the second procurement description to obtain the second content, performs text segmentation on the second content to obtain the second segmentation result, and constructs a second word vector based on the second segmentation result, wherein the second word vector includes each segmentation and the word frequency of each segmentation. S103 calculates the cosine similarity or simple common word similarity between the first content word vector and the second content word vector, and uses the calculation result as the procurement content similarity.

[0030] Among them, the procurement description can be a relatively important document that records the procurement characteristics in the procurement process. Specifically, the document recording the procurement characteristics may include a procurement bidding document or a procurement demand document. The first procurement description corresponds to the first procurement process, and the second procurement description corresponds to the second procurement process. Regardless of whether the procurement process is successfully tendered, the relevant documents can be collected and analyzed according to this solution. This solution analyzes and compares the word vectors that appear in the procurement description. The closer the word vectors are, the higher the similarity of the procurement-related documents.

[0031] The above technical solution can compare the similarity of different purchase contents through the analysis and comparison of the word vector attributes of the set purchase instructions, can quickly determine the correlation between multiple purchase processes, and provide a reference for evaluating the purchase system. In some embodiments of the purchase data management system, when collecting purchase information, when calculating the purchase information to be collected, the new purchase information can be used as a node, and the purchase content similarity of the new purchase information and all existing purchase information nodes can be stored, and a storage data relationship between nodes connected by purchase content similarity can be established.

[0032] In some other specific embodiments, the purchase similarity value calculation module 102 is used to select the core content of the purchase description, including the steps of:

[0033] Identify the templated content in the procurement instructions, including the tender notice, tender instructions, contract terms, and tender document format requirements, delete the templated content, and obtain the first content. Identifying the templated content in the procurement instructions can be achieved by identifying paragraph keywords. The applicant found that in actual applications, the templated content of the procurement instructions is similar, so there are many similar word structures. Including templated content in the evaluation system may cause the similarity value to increase inappropriately. Therefore, by identifying and deleting the woodblock print content and selecting the core content, it is possible to better eliminate the influence of the templated content in the procurement instructions on the similarity judgment, and achieve a better technical effect of evaluating the similarity.

[0034] In some other specific embodiments, in order to further eliminate the noise in the word segmentation results, the applicant further sets the following scheme:

[0035] The purchase similarity value calculation module 102 performs a step of segmenting the first content and tagging the part of speech, and can also calculate the first noun vector V tn and the first verb vector V tv and the first other word vector V to , filter stop words and common business words; after the second content is segmented, the step of part-of-speech tagging is performed to obtain the second noun vector V tn and the second verb vector V tv And the second other word vector V to , filter stop words and common business words. In this embodiment, part-of-speech tagging helps to distinguish the part-of-speech of each word. At the same time, filtering stop words and common business words can also avoid the influence of similar business words on similarity judgment. Common business words in some embodiments, such as common words for software information construction projects, include: high cohesion, low coupling, microservices, etc. Common words in the business field can be manually annotated. In this way, noise can be eliminated and the similarity of different procurement instructions can be better compared.

[0036] In some other specific embodiments, the procurement similarity value calculation module 102 is further used to execute the steps of increasing the weight of the second noun vector and the second verb vector in the first content word vector; increasing the weight of the second noun vector and the second verb vector in the second content word vector. The core content of procurement is unique in that it contains many nouns and verbs, so increasing the weight of nouns and verbs helps to distinguish different procurement description contents, thereby making the judgment of procurement similarity more scientific and effective.

[0037] In some other specific embodiments, the procurement similarity value calculation module 102 is also used to execute the steps of using regular expressions to identify titles for the first content, wherein the titles include large titles, small titles, main titles, and subtitles; performing word segmentation and part-of-speech tagging on the titles to form a first title sentence word vector, and increasing the weight of the first title sentence word vector in the first content word vector; using regular expressions to identify titles for the second content, wherein the titles include large titles, small titles, main titles, and subtitles; performing word segmentation and part-of-speech tagging on the titles to form a second title sentence word vector, and increasing the weight of the second title sentence word vector in the first content word vector. Title identification helps to locate and control the main content of the procurement instructions. Recognizing text titles through regular expressions can be more accurate and efficient. After identifying the title, the weight of the word vector contained in the title can be increased, which can better highlight the key word segmentation in the procurement instructions, making the similarity judgment more scientific, and the similarity score can better reflect the similarity of different procurement instructions.

[0038] In some other embodiments, the purchase similarity value calculation module 102 is further configured to perform the following steps:

[0039] Eliminate templated content in bidding documents, such as: bidding announcement, bidding instructions, contract terms, bidding document format requirements, bid evaluation methods, etc.

[0040] Procurement content processing in the bidding document uses regular expressions to identify the title of each sentence and form a title list T l Then the title sentence is segmented and POS tagged to form a noun vector V tn and the verb vector V tv and other word vectors V to ,The dimension of the vector is the union of non-repeated words, and the weight of each dimension is the word frequency.

[0041] Perform word segmentation and part-of-speech tagging on each sentence in the procurement content of the bidding document to form a noun vector V cn and the verb vector V cv and other word vectors V co ,The dimension of the vector is the union of non-repeated words, and the weight of each dimension is the word frequency.

[0042] Filter out stop words.

[0043] Filter out common words in the business field, such as common words in software information construction projects: high cohesion, low coupling, microservices, etc. Common words in the business field can be manually annotated.

[0044] Improve the title sentence segmentation to form the noun vector V tn and the verb vector V tv and other word vectors V to The weight of each dimension is the word frequency multiplied by the weighting coefficient Δt (Δt>1, the title is more important than the text content).

[0045] Combine the word vectors formed by the title and other content in the text and add the weight values ​​of each dimension of the same word to get:

[0046] Noun word vector: Vn = V tn ∪V cn

[0047] Verb word vector: Vt = V tv ∪V cv

[0048] Other word vectors: Vo = V to ∪V co

[0049] Improve the noun word vector Vn and the verb word vector V t The weight of each dimension in the noun word vector V n The weight value of each dimension is the word frequency multiplied by Δn, the verb word vector V t The weight value of each dimension is the word frequency multiplied by Δv (Δn>1, Δv>1, the core content of procurement is mostly nouns and verbs, so the weights of nouns and verbs are increased).

[0050] Merge noun word vector V n , verb word vector V t Together with other word vectors Vo, the final procurement bidding document word vector V is formed.

[0051] The bidding documents of the two purchases to be compared are processed according to the above steps to obtain the two purchase word vectors, which are recorded as: V i and V j . Calculate the cosine similarity of word vectors or simple common words as similarity C(i,j).

[0052] In some other specific embodiments, the storage module 101 is also used to store enterprise information and the connection between procurement information and participating enterprises. By setting up a storage module to store the connection between procurement information and participating enterprises, the node network in the procurement data management system can be made more perfect, and when calculating enterprise information, relevant procurement information and procurement content similarity can be called for calculation, and enterprises can also be connected in series through procurement information.

[0053] In some other specific embodiments, after receiving new procurement information, the storage module 101 calculates the procurement content similarity between the new procurement information and the stored procurement information if the participating enterprises of the new procurement information are the same as at least one participating enterprise of the stored procurement information, and saves the procurement content similarity between the new procurement information and the stored procurement information. For example, the storage module stores procurement information a and b, wherein the participating enterprises of procurement information a are A, B, and C, and the participating enterprises of procurement information b are D, E, and F. At this time, it is not necessary to calculate the procurement content similarity to connect a and b. Only when procurement information c is newly entered, the storage system retrieves that the participating enterprises of procurement information c are C, G, and H. At this time, it is necessary to establish the connection between procurement information c and a. Therefore, the procurement content similarity between procurement information c and procurement information a is calculated, and it is still not necessary to calculate and store the procurement content similarity between procurement information c and procurement information b. The above-mentioned calculation of procurement content similarity is only calculated and stored between procurement information that is connected through participating enterprises, which saves the amount of calculation and data storage required by the data management system, making the operation of the procurement data management system more efficient.

[0054] In some other embodiments, after receiving new purchase information, the storage module calculates the purchase content similarity between the new purchase information and the stored purchase information, and saves the purchase content similarity between the new purchase information and the stored purchase information. Recording the content similarity between the new purchase information and all the stored purchase information can record the purchase information more completely and specifically.

[0055] In some other further embodiments, the procurement data management system further includes a supplier scoring module 103, which is used to calculate the supplier score of the enterprise relative to each procurement information in which it participates according to preset rules. The preset rules can be determined according to user needs. For example, the degree of fit between the supplier and the specific procurement information is calculated based on the similarity of the procurement content of the procurement information in which the supplier enterprise has participated. By scoring the supplier relative to the specific procurement information, the degree of fit between the supplier and the specific procurement information can be better demonstrated, which helps to improve the sunshine and transparency of procurement data.

[0056] In some further embodiments, the storage module 101 is also used to store supplier ratings between the enterprise and the procurement information that the enterprise participates in. Storing supplier ratings can make the data of the procurement data management system more complete.

[0057] In some other embodiments, the supplier scoring module 103 also performs the following operations for the user: S1 calculates the similarity of procurement content; S2 calculates the similarity of supplier services; S3 calculates the supplier business score; and S4 calculates the matching degree of supplier bid quotations.

[0058] S4 calculates the total score of the first supplier, and the total score of the first supplier is positively correlated with the procurement content similarity, the supplier service content similarity, the supplier business score, and the supplier bid quotation matching degree.

[0059] Among them, the procurement content similarity is the similarity between the supplier's historical procurement content and the current procurement content, the supplier service similarity is the value obtained by considering the time factor of the supplier's historical procurement content, and the supplier quotation matching is the value obtained by considering the procurement content similarity of the supplier's quotation. In some embodiments, the supplier's score Ai can be set to SCi*ΔC+ECi*ΔE+PCi*ΔP.

[0060] All suppliers participating in this bidding are traversed, and the supplier procurement content similarity SCi, supplier business score ECi, and supplier bid quotation matching degree PCi are calculated respectively. ΔC, ΔE, and ΔP are weight values.

[0061] The above scheme can determine whether a supplier has relevant cooperation value from multiple dimensions by calculating the service similarity of the supplier, calculating the supplier's business score, and calculating the matching degree of the supplier's bid quotation. The above scheme proposes a supplier scoring method, which can improve the degree of automation, provide a reference for users to select purchasers, and improve the program's degree of automation and fairness in supplier selection.

[0062] The method for calculating supplier service similarity by the supplier scoring module 103 includes the following steps:

[0063] The similarity of the single purchase content of all the bids in which the first supplier has participated is obtained, and the single purchase content similarity of all the bids is multiplied by the time correction factor of all the bids to obtain the supplier service similarity, and the time correction factor of all the bids is positively correlated with the time of all the bids. In some specific embodiments, the calculation method of the supplier service similarity can be:

[0064] 1) Initialize the supplier procurement content similarity SC to 0.

[0065] 2) Traverse each bid in which the supplier has participated and calculate the similarity of the content of a single purchase, recorded as Ci, and calculate the time correction factor Ti for each purchase.

[0066] 3) Accumulate the similarity of supplier procurement content, SC = SC + Ci*Ti.

[0067] 4) The final accumulated value is the similarity of SC supplier procurement content.

[0068] The time correction factor is positively correlated with the time of each bid. The closer the time of a bid is to the current time, the greater the value of the time correction factor. Considering that the bid content, supplier qualifications, etc. will change over time, and their reference value in the bidding process will become weaker and weaker as time goes by, a time correction factor T is introduced for entities with time factors. In some embodiments, the time correction factor T is defined as follows:

[0069] 1) The base time is defined as 00:00:00 on January 1, 1970.

[0070] 2) Calculate the time difference ΔTa between the bid time of historical bid A and the reference time, in seconds.

[0071] 3) Calculate the time difference ΔT between this bidding time and the reference time in seconds.

[0072] 4) Then T = ΔTa / ΔT.

[0073] The reference time can be set as needed. In other specific embodiments,

[0074] The time correction factor Ti for the i-th bidding satisfies:

[0075] Ti=(ti-t0) / (tc-t0)

[0076] ti is the i-th bidding time, tc is the current bidding time, and t0 is the reference time constant. Through the above time correction factor, the procurement process closer to the current time can have a higher impact value, thereby better solving the practical problem of supplier scoring.

[0077] In some further specific embodiments, the method for calculating the supplier bid price matching degree by the supplier scoring module 103 includes the following steps:

[0078] Obtain the similarity of the contents of all previous bids and purchases in which the first supplier has participated, obtain the bid quotations of all previous bids and purchases in which the first supplier has participated, obtain the bid quotation matching degrees of all previous bids according to the formula: bid quotation matching degree = bid procurement content similarity / bid quotation, multiply the bid quotation matching degrees of all previous bids by the time correction factors of all previous bids, and accumulate the results to obtain the supplier bid price matching degree, wherein the time correction factors of all previous bids are positively correlated with the time of occurrence of all previous bids.

[0079] For example, in some embodiments, the scheme calculates the bid matching degree:

[0080] The calculation algorithm for the matching degree of a single bidding quotation is as follows:

[0081] 1) Get the bid price Pi of the supplier

[0082] 2) Calculate the similarity Ci of single purchase content.

[0083] 3) Calculate the single bid quotation matching degree ΔP based on the similarity of the procurement content, and the calculation formula is: ΔP = Ci / Pi.

[0084] The supplier bid quotation matching calculation algorithm is as follows:

[0085] 1) Initialize the supplier's bid quotation matching degree PC to 0.

[0086] 2) Traverse each bid in which the supplier participates and calculate the single quotation matching degree ΔPi, and at the same time calculate the time correction factor Ti for each purchase.

[0087] 3) Accumulate the supplier's bid quotation matching degree PC = PC + ΔPi*Ti. Complete the accumulation of the results of all single quotation matching degrees multiplied by the time correction factor. Through the above settings, the procurement process closer to the current time can have a higher impact value, thereby quantifying the supplier's bid quotation matching degree, thereby better solving the problem of the practicality of supplier scoring.

[0088] In some other specific embodiments, the present solution further performs the steps of calculating the bid business score evaluation and calculating the supplier's bid business score evaluation algorithm as follows:

[0089] 1) Initialize the supplier's business evaluation value EC to 0.

[0090] 2) Traverse each bid that the supplier has participated in and obtain the business score value. The business score value here refers to the business score value obtained by the enterprise after participating in the bidding evaluation and scoring according to the method specified in the business score in the bid document. If it is not calculated in percentage, it is normalized to percentage, recorded as Ei, and the time adjustment factor Ti of each purchase is calculated at the same time.

[0091] 3) Accumulate the similarity of supplier procurement content, EC = EC + Ci*Ti.

[0092] 4) The final accumulated value EC is the bid commercial evaluation value. Through the above settings, the procurement process closer to the current time can have a higher impact value, thereby quantifying the current commercial score of the supplier, thereby better solving the practicality problem of supplier scoring.

[0093] It should be noted that, although the above embodiments have been described in this article, the patent protection scope of the present invention is not limited thereby. Therefore, based on the innovative concept of the present invention, changes and modifications made to the embodiments described herein, or equivalent structures or equivalent process changes made using the contents of the present invention specification and drawings, directly or indirectly applying the above technical solutions to other related technical fields, are all included in the patent protection scope of the present invention.

Claims

1. A procurement big data management system, characterized in that: Includes the following modules: a storage module, the storage module is used to store procurement information, the procurement information includes procurement instructions, and is also used to store procurement content similarity between at least two procurement information, A procurement similarity value calculation module, the procurement similarity value calculation module is used to execute the steps of selecting the core content of the first procurement description to obtain the first content, performing text segmentation on the first content to obtain the first segmentation result, constructing a first content word vector according to the first segmentation result, the first content word vector includes each segmentation and the word frequency of each segmentation, selecting the core content of the second procurement description to obtain the second content, performing text segmentation on the second content to obtain the second segmentation result, constructing a second content word vector according to the second segmentation result, the second content word vector includes each segmentation and the word frequency of each segmentation, The purchase similarity value calculation module is used to calculate the purchase content similarity based on the cosine similarity or simple common word similarity between the first content word vector and the second content word vector; The storage module is also used to store enterprise information and the connection between procurement information and participating enterprises; After receiving new procurement information, the storage module calculates the procurement content similarity between the new procurement information and the stored procurement information only if the participating enterprise of the new procurement information is the same as at least one participating enterprise of the stored procurement information, and saves the procurement content similarity between the new procurement information and the stored procurement information.

2. The procurement big data management system according to claim 1, characterized in that: The procurement similarity value calculation module is used to select the core content of the procurement instructions, and specifically includes the steps of identifying the templated content in the procurement instructions, including the tender notice, tender instructions, contract terms, and tender document format requirements, deleting the templated content, and obtaining the selected content.

3. The procurement big data management system according to claim 1, characterized in that: Perform part-of-speech tagging and increase the weights of the noun vectors and verb vectors in the first content word vector and the second content word vector.

4. The procurement big data management system according to claim 1, characterized in that: It also includes a title recognition module, which is used to use regular expressions to recognize titles of the first content and the second content, wherein the titles include main titles, subtitles, main titles, and subtitles; perform word segmentation and part-of-speech tagging on the titles, The purchase similarity value calculation module is used to add weight to the word vector obtained by the title sentence segmentation in the first content word vector; and is also used to add weight to the title sentence segmentation vector in the second content word vector.

5. The procurement big data management system according to claim 1, characterized in that: It also includes a supplier scoring module, which is used to calculate the supplier scores of the enterprise relative to the various procurement information in which it participates according to preset rules.

6. The procurement big data management system according to claim 5, characterized in that: The storage module is also used to store supplier ratings between the enterprise and the procurement information in which the enterprise participates.

Images