CN111242679A - Sales forecasting method based on product review viewpoint mining - Google Patents
Sales forecasting method based on product review viewpoint mining Download PDFInfo
- Publication number
- CN111242679A CN111242679A CN202010017679.0A CN202010017679A CN111242679A CN 111242679 A CN111242679 A CN 111242679A CN 202010017679 A CN202010017679 A CN 202010017679A CN 111242679 A CN111242679 A CN 111242679A
- Authority
- CN
- China
- Prior art keywords
- product
- attribute
- product attribute
- comments
- sales
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3335—Syntactic pre-processing, e.g. stopword elimination, stemming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/374—Thesaurus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
Abstract
The invention discloses a sales prediction method based on product comment viewpoint mining, which is used for performing aspect-level viewpoint mining on initial comments and additional comments of a product, extracting attributes and corresponding evaluation information of the product, analyzing the emotional polarity of each attribute of the product by a consumer through an emotion classification model, further performing emotion classification and quantization on comment texts, merging a quantization value with early-stage historical sales into a sales prediction model to predict product sales, and improving prediction capability and accuracy. According to the method, additional comments are introduced on the basis of the initial comments, the initial comments and the additional comments are subjected to aspect-level viewpoint mining, products are analyzed in a multi-dimensional mode, sentiment analysis is carried out on different attributes, comment texts are quantized to obtain scores of the different attributes of the products, the scores are combined with historical sales to predict future sales of the products, and the accuracy of sales prediction is improved.
Description
Technical Field
The invention belongs to the technical field of natural language processing, and particularly relates to a sales forecasting method based on product review viewpoint mining.
Background
With the rapid development of electronic commerce, the behavior patterns of consumers are changed greatly, and online shopping becomes a hot tide. Compared with the traditional shopping, the online shopping has the characteristics of diversity, low shopping cost, no time and space limitation and the like, great convenience is brought to people, the online shopping gradually becomes a consumption habit of people, more and more consumers like purchasing products on E-commerce websites (such as Jingdong, Taobao, Amazon and the like), and then shopping evaluation is made in an online comment mode, the online comment represents the emotional polarity of the consumers on the products, and the online comment mainly contains valuable information such as views, satisfaction, opinions and the like of different attributes of the products. The merchant knows the real demand of the consumer through the online comment, improves the product, improves the service, adjusts the sale strategy and improves the sales volume of the product; and the consumer obtains the public praise of the product through the comment and makes a purchase decision.
According to statistics, most consumers can browse online reviews of products besides paying attention to the properties of the products when purchasing the products. Compared with product information issued by merchants, the product reviews have more obvious credibility and persuasion, are main reference information sources of consumer online shopping, and directly influence shopping willingness and decision making of the consumers, and further influence sales volume of products. Therefore, the emotional factors of the product comments are mined aiming at different product attributes, so that the real emotional expression of the consumer can be more approximate, and the analysis of which attributes of the product can influence the product sales volume, thereby improving the sales volume prediction capability and accuracy and providing application reference for the decision of the merchant.
1. Situation of view mining
The viewpoint mining is also called emotion analysis, and is used for mining and analyzing emotion information such as the theme, subjectivity and emotion attitude of text information and further identifying the emotional tendency of subjective text. The research objects are mainly texts on the Web, and particularly comment texts published by users. According to the granularity of the analyzed text, the viewpoint mining can be divided into three categories of chapter level, sentence level and aspect level. There are major categories of opinion mining based on product reviews and news reviews, according to the category of the analyzed text. The methods of viewpoint mining mainly involve three types:
(1) based on the emotion dictionary: the traditional method for judging the emotion polarity of subjective text is to construct an emotion dictionary based on emotion knowledge and use the emotion dictionary as a tool. And (3) carrying out induction and arrangement on the widely used emotion words according to experience, matching the text to be processed with the words in the emotion dictionary, and counting the occurrence times of the positive and negative emotion words so as to judge the emotion polarity of the text. The method mainly depends on the construction of an emotion dictionary, has certain limitation, cannot cover new words, and enables the text emotion judgment accuracy to be low.
(2) Based on machine learning: after text features are manually extracted, the computer processes the text according to a certain machine learning algorithm and outputs emotion classification. Common machine learning methods include naive Bayes NB, support vector machine SVM, and maximum entropy model. The machine learning method mainly solves the problems that the representation of good texts, the selection of characteristics and a classifier need to manually mark text characteristics, the workload is large, and in addition, the method cannot learn deep semantics which are usually shallow learning.
(3) Based on deep learning: the deep learning can analyze the text, automatically extract the features and automatically learn and optimize the model output by constructing a neural network model, so that the work of manually extracting the features is reduced, the neural network can learn the semantic information of the context, and the emotion classification accuracy is improved. Common network models are convolutional neural networks CNN, long short term memory networks LSTM, gate control units GRU, and variants thereof.
2. Status quo of sales
From the selection of research variables, the input of sales predictions can be divided into both structured and unstructured data. The structured data comprises historical sales, comment star level, initial comment quantity, good comment rate, comment length and the like which are used as input of a regression model to predict sales. The unstructured data are first comment texts, the texts are digitalized through viewpoint mining, the sales volume is predicted by fusing the structured data, and common sales volume prediction models comprise ARAM models, SVM models, LSTM models and the like. The study of the scholars has been mainly focused on the first comments, and few scholars consider to add comments. The product is not deeply used by a consumer when the initial comment is published, the internal quality of the product is ignored only because external factors such as product appearance and packaging give good comments, and for the additional comment, the user personally experiences the product and more objectively reflects the real feeling of the user, and product attributes which do not appear in the initial comment often appear in the additional comment, so that the initial comment is effectively supplemented and corrected, more comprehensive, more useful and more reliable information can be brought, the current hotspot problem can be better reflected, the influence on the purchasing decision of the consumer is more profound, and the sales volume of the product can be further influenced.
The method introduces additional comments on the basis of the initial comments, performs aspect-level viewpoint mining on the initial comments and the additional comments, performs multidimensional product analysis and sentiment analysis on different attributes, further quantizes comment texts to obtain scores of the different attributes of the product, and predicts the future sales volume of the product by combining the score values with the historical sales volume.
Disclosure of Invention
The invention aims to provide a sales prediction method based on product comment viewpoint mining, which is characterized in that the first comment and the additional comment of a product are subjected to aspect-level viewpoint mining, the attribute of the product and corresponding evaluation information are extracted, the emotion polarity of each attribute of the product by a consumer is analyzed through an emotion classification model, then comment texts are subjected to emotion classification and quantization, a quantization value is combined with the historical sales in the previous stage to predict the product sales, and the prediction capability and accuracy are improved.
The technical scheme adopted by the invention is a sales forecasting method based on product review viewpoint mining, which mainly comprises the following steps:
step (1), data acquisition: and (3) crawling product comments, including initial comments and additional comments, on E-commerce websites such as Taobao and Jingdong.
Step (2) data preprocessing: firstly, removing repeated and short comments from the collected comments; secondly, loading the initial comments and the additional comments to a word segmentation toolkit, and segmenting words of the text by using a word segmentation tool; then, loading a stop word list to remove stop words; and finally, training word vectors by using the processed word segmentation corpus, and converting the text into vector representation.
Extracting product attribute viewpoint pairs: after the comment text is converted into vector representation, extracting the product attributes of each initial comment and each additional comment and the corresponding emotional intensity and emotional word by using a sequence tagging model, expressing the product attributes and the additional comment into a triple form (product attributes, emotional words and degree adverbs), and constructing a product attribute viewpoint co-occurrence matrix.
Clustering the extracted product attributes: different consumers have different expression modes for the same product attribute, so that synonym clustering needs to be carried out on the product attributes extracted in the step (3), and one attribute in the synonyms is selected as the evaluated product attribute.
And (5) judging the emotional polarity of the product attribute: inputting the text converted into the vector into an emotion classification model to identify the emotion polarity of each product attribute in the initial comment and the additional comment, wherein the emotion polarities are divided into three types of positive direction, neutral direction and negative direction, which are respectively represented by 1, 0 and-1, and the model prediction result is stored in a binary form (product attribute and emotion polarity), wherein the product attribute is the product attribute extracted in the step (4).
Step (6) calculating a product attribute score value: and (5) calculating the sentiment value of the product attribute according to the sentiment polarity of the product attribute identified in the step (5), and calculating the score value of the product attribute according to the sentiment value of the product attribute. The value range of the product attribute emotion value is [ -1,1], the value range of the product attribute score value is [1,5], and in order to correspond to the value range of the product attribute score value, the product attribute emotion value x can be converted into the product attribute score value y through a mapping function y of 2x + 3.
Step (7) calculating the total product score value: and (4) acquiring the occurrence frequency of each attribute of the product in the co-occurrence matrix in the step (3), calculating the weight of the product attribute according to the frequency, and linearly adding the weight of the product attribute and the product attribute score value calculated in the step (6) to obtain the total score value of the product.
And (8) predicting sales volume: and (4) according to the operations from the step (1) to the step (7), taking the total product score value calculated by the primary comment and the total product score value calculated by the additional comment as input of a sales prediction model by combining the historical sales in the early stage, training the sales prediction model, and predicting the future sales.
The previous research mainly aims at the first comment, the consumer probably does not deeply use the product when the first comment is published, the internal quality of the product is ignored, the additional comment is an opinion published after the user personally experiences the product and more objectively reflects the real feeling of the user, the product attribute which does not appear in the first comment often appears in the additional comment, the additional comment is effective supplement and correction for the first comment, more comprehensive, more useful and more reliable information can be brought, the current hotspot problem can be reflected, the purchase decision of the potential consumer can be influenced, and the sales volume of the product can be further influenced. According to the method, additional comments are introduced on the basis of the initial comments, fine-grained aspect-level viewpoint mining is performed on the initial comments and the additional comments, the comment texts are quantized to obtain scores of different attributes of the product, the future sales volume of the product is predicted by combining the score values with the historical sales volume, and the accuracy of sales volume prediction is improved.
Drawings
FIG. 1 is an overall flow chart provided by the present invention;
FIG. 2 is a diagram of a product attribute viewpoint pair extraction model in accordance with the present invention;
FIG. 3 is a diagram of the structure of the BilSTM-CRF model in the present invention;
FIG. 4 is a sample word sequence tagging of the present invention;
FIG. 5 is a sample diagram of a co-occurrence matrix of product attribute viewpoint pairs as contemplated by the present invention;
FIG. 6 is an illustration of a sample glossary of degree adverbs for the design of the present invention.
Detailed Description
The implementation steps of the invention are described in detail in the attached drawings of the specification. A sales forecasting method based on product review viewpoint mining comprises the following steps:
s1, data acquisition and cleaning: product comments are crawled on the E-commerce websites of Taobao, Jingdong and the like, the product comments comprise initial comments and additional comments, repeated comments and short comments are deleted, the longer the comment is, the more perfect the introduction of the product is, the more product attributes are contained, the larger the amount of information available for other users to buy, and the clearer and more accurate the purchasing decision is.
S2, clause: the cleaned initial comment and the additional comment are divided into a plurality of short sentences according to the punctuations such as commas, semicolons, periods and the like, and each short sentence usually only contains one attribute of the product, so that the method is favorable for counting the co-occurrence times of the product attribute and the viewpoint word in the step S6 and judging the emotional polarity of the product attribute in the step S8. Suppose the format of a comment is: "sen1, sen2, sen3", the result of punctuation into multiple short sentences is: [ sen1, sen2, sen3]
S3, word segmentation: using the jieba word segmentation tool to segment each short sentence in the comment, for the example in step S2, the result after segmentation is: [ [ w1, w2, w3], [ w1, w2], [ w1, w2, w3, w4] ] ].
S4, removing stop words: the words of chinese text are generally divided into real words and imaginary words. The real words are words with practical meaning, while the imaginary words are words without practical meaning, such as' true, false, and, etc., and the imaginary words do not work in the viewpoint mining and may generate noise, so that the imaginary words are filtered out, and the stop words (imaginary words) in the short sentences after word segmentation can be deleted according to the existing stop word list, such as the Hadoop stop word list and the Baidu stop word list.
S5, generating a word vector: the text processed in S4 is converted to a numerical type and converted to a word vector representation by google' S word2vec tool. The word vector retains semantic information of the word, and facilitates processing of S7 and S8.
S6, extracting product attribute viewpoint pairs: product attributes are divided into explicit and implicit attributes. The explicit attribute directly indicates the product attribute corresponding to the viewpoint in the comment, for example, the pixel in the 'good' cell phone pixel is the explicit attribute, the 'good' is the adverb, and the 'good' is the viewpoint word. Implicit attributes refer to which attributes of a product are not described in a review, but product attributes can be inferred through certain terms or semantics, such as "price" which is an implicit attribute described in "mobile phone is expensive," but the term "price" does not appear in the review.
The extraction of the product attribute viewpoint pair comprises two aspects:
(1) and (3) extracting the explicit attribute viewpoint pairs: it can be regarded as a sequence labeling problem, and the labeling set adopts a BIO mode. Each word in the comment corresponds to one label, 7 labels are provided in total, B-feature represents the head word of the product attribute, I-feature represents the word after the head word of the product attribute, B-depth represents the head word of the degree word, I-depth represents the word after the head word of the degree word, B-opinion represents the head word of the viewpoint word, I-opinion represents the word after the head word of the viewpoint word, and O represents other words. The comment callout is shown, for example, in fig. 4. The sequence labeling model can adopt a BilSTM _ CRF model, and a softmax layer of the BilSTM model is replaced by the CRF, so that the training of the model becomes an end-to-end process, the method does not depend on feature engineering, BilSTM can learn the semantic information of the context, and the CRF considers the dependency relationship among labels, so that the output of the current label not only depends on the current input, but also considers the output label at the previous moment, and the problem of label bias is solved, therefore, the text vector obtained in the step S5 is used as the input, and a good effect can be achieved.
The structure of the BilSTM-CRF model is shown in figure 3.
(2) Extracting implicit attribute viewpoint pairs: an attribute and viewpoint co-occurrence matrix is constructed for the extracted explicit attributes and viewpoint words, rows of the matrix represent product attributes, columns represent viewpoint words, values in the matrix represent co-occurrence times of the product attributes and the viewpoint words, and a sample diagram of the co-occurrence matrix is shown in fig. 5. Extracting the viewpoint words s in the comment, calculating the explicit attribute f in the product attribute viewpoint co-occurrence matrix and the collocation weight of the viewpoint words s by using an improved tf-idf algorithm, selecting the attribute f corresponding to the maximum collocation weight as the implicit attribute corresponding to the viewpoint words s in the comment, and adding the implicit attribute and the viewpoint words into the co-occurrence matrix, thereby facilitating the extraction of the subsequent implicit attribute. The extracted product attribute viewpoint pairs are stored in a triple form (product attributes, emotional words and degree adverbs). The improved tf-idf is calculated as follows:
wherein: f is the product attribute in the co-occurrence matrix, s is the viewpoint word, A is all the attributes of the product in the co-occurrence matrix, w (f, s) is the collocation weight of the product attribute f and the viewpoint word s, freq (f, s) is the co-occurrence frequency of a certain attribute f of the product and the viewpoint word s, freq (A, s) is the co-occurrence frequency of the viewpoint word s and each attribute of the product, n (A) is the number of the product attribute in the co-occurrence matrix, and n (A, s) is the number of the product attribute co-occurring with the viewpoint word s.
S7, clustering product attributes: different consumers may express the same product attribute differently, for example, the "appearance" expresses the same attribute, and the "appearance" can be expressed by the term. The synonym clustering in the step S6 is realized by calculating semantic similarity, one product attribute is selected to replace other synonyms, and the product attributes in the product attribute viewpoint co-occurrence matrix in the step S6 are also changed correspondingly. The calculation formula of the similarity is as follows:
wherein: a and b are word vectors of two words, N represents the dimension of the word vector
S8, judging the emotion polarity of the product attribute: converting the short sentence of each first comment and each additional comment after word segmentation into a vector, inputting the vector into an emotion classification model, and judging the emotion polarity of the short sentence. In the method, the emotion types are divided into three types of positive direction, neutral direction and negative direction which are respectively represented by 1, 0 and-1, and the prediction result of each short sentence is stored in a binary form (product attribute and emotion polarity), wherein the product attribute is the product attribute extracted in S6.
S9, calculating the product score value: the score values of the products are calculated from the primary and additional reviews, respectively, and are represented by first _ score (product) and second _ score (product), respectively. S9 includes two parts:
(1) calculating the score value of the product attribute: and calculating the score value of the product attribute according to the emotion polarity of the product attribute obtained in the step S8, wherein the value range of the emotion value of the product attribute is [ -1,1], and in order to correspond to the value range [1,5] of the product attribute score value, the emotion value x of the product attribute can be converted into the score value y of the product attribute by a mapping function y of 2x + 3. The scoring formula for the product attributes is as follows:
wherein: socre (f) is the value of the product attribute f, s (f)i) For the emotional polarity of the product attribute f in each comment, w (adv) is the weight of the degree adverb, a sample diagram of the weight of the degree adverb is shown in fig. 6, and n (f) is the number of the product attribute f, which is obtained by the co-occurrence matrix in S6.
(2) Calculating the overall score value of the product: counting the occurrence times of each attribute of the product according to the co-occurrence matrix of S6, and calculating the weight of each attribute through a formula 4; and calculating the overall product score value according to the product attribute weight and the product attribute score value obtained by the formula 3 by using a formula 5.
Wherein: n (f) is the total number of times the product attribute f appears in the first review or additional reviews,
is the total number of times of all product attributes in the initial review or additional review, and m is the number of categories of product attributes.
Wherein: w (f)i) Is the weight of the product attribute f in the first comment or additional comment, score (f)i) The value of the product attribute f in the first comment or additional comment is shown, and M is the total number of the first comment or additional comment.
S10, sales prediction: calculating the product score of the first comment and the product score of the additional comment in each month according to the flow from S1 to S9 in a month period, and predicting the sales of the next month of the merchant by combining the previous monthly sales as the input characteristics of a sales prediction model, wherein the sales prediction regression model has the following expression:
wherein: sale (product) represents sales in the next month, FiProduct rating, S, representing first commentiProduct score, Q, representing additional reviewsiRepresents historical monthly sales, T represents data taken T months before,
βi、Qiare respectively Fi、Si、QiU represents a constant.
Claims (3)
1. A sales forecasting method based on product review viewpoint mining is characterized by comprising the following steps: the method comprises the following steps:
step (1), data acquisition: crawling product comments on the E-commerce website, wherein the product comments comprise initial comments and additional comments;
step (2) data preprocessing: firstly, removing repeated and short comments from the collected comments; secondly, loading the initial comments and the additional comments to a word segmentation toolkit, and segmenting words of the text by using a word segmentation tool; then, loading a stop word list to remove stop words; finally, training word vectors by using the processed word segmentation corpus, and converting the text into vector representation;
extracting product attribute viewpoint pairs: after the comment text is converted into vector representation, extracting the product attributes of each initial comment and each additional comment and corresponding emotional intensity and emotional words by using a sequence tagging model, expressing the product attributes and the added comments into a triple form, and constructing a product attribute viewpoint co-occurrence matrix;
clustering the extracted product attributes: different consumers have different expression modes for the same product attribute, so that synonym clustering needs to be carried out on the product attribute extracted in the step (3), and one attribute in synonyms is selected as the evaluated product attribute;
and (5) judging the emotional polarity of the product attribute: inputting the text converted into the vector into an emotion classification model to identify the emotion polarity of each product attribute in the initial comment and the additional comment, wherein the emotion polarity is divided into three types of positive direction, neutral direction and negative direction, which are respectively represented by 1, 0 and-1, and the model prediction result is stored in a binary form, wherein the product attribute is the product attribute extracted in the step (4);
step (6) calculating a product attribute score value: calculating the sentiment value of the product attribute according to the sentiment polarity of the product attribute identified in the step (5), and calculating the score value of the product attribute according to the sentiment value of the product attribute; the value range of the product attribute emotion value is [ -1,1], the value range of the product attribute score value is [1,5], and in order to correspond to the value range of the product attribute score value, the product attribute emotion value x can be converted into the product attribute score value y through a mapping function y of 2x + 3;
step (7) calculating the total product score value: acquiring the occurrence frequency of each attribute of the product in the co-occurrence matrix in the step (3), calculating the weight of the product attribute according to the frequency, and linearly adding the weight of the product attribute and the product attribute score value calculated in the step (6) to obtain the total score value of the product;
and (8) predicting sales volume: and (4) according to the operations from the step (1) to the step (7), taking the total product score value calculated by the primary comment and the total product score value calculated by the additional comment as input of a sales prediction model by combining the historical sales in the early stage, training the sales prediction model, and predicting the future sales.
2. The sales forecasting method mined based on the product review viewpoint according to claim 1, characterized in that: the triple form comprises product attributes, emotion words and degree adverbs.
3. The sales forecasting method mined based on the product review viewpoint according to claim 1, characterized in that: the binary form comprises product attributes and emotion polarities.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010017679.0A CN111242679A (en) | 2020-01-08 | 2020-01-08 | Sales forecasting method based on product review viewpoint mining |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010017679.0A CN111242679A (en) | 2020-01-08 | 2020-01-08 | Sales forecasting method based on product review viewpoint mining |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111242679A true CN111242679A (en) | 2020-06-05 |
Family
ID=70874369
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010017679.0A Pending CN111242679A (en) | 2020-01-08 | 2020-01-08 | Sales forecasting method based on product review viewpoint mining |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111242679A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991767A (en) * | 2019-12-20 | 2020-04-10 | 浙江大学 | Leading user identification and prediction method and technical trend prediction method |
| CN111882039A (en) * | 2020-07-28 | 2020-11-03 | 平安科技（深圳）有限公司 | Physical machine sales data prediction method and device, computer equipment and storage medium |
| CN112069402A (en) * | 2020-08-28 | 2020-12-11 | 山东省人工智能研究院 | Personalized comment recommendation method based on emotion and graph convolution neural network |
| CN113393279A (en) * | 2021-07-08 | 2021-09-14 | 北京沃东天骏信息技术有限公司 | Order quantity estimation method and system |
| CN113516164A (en) * | 2021-04-30 | 2021-10-19 | 辽宁工程技术大学 | Fruit tree disease and insect pest diagnosis method integrating knowledge map and deep learning |
| CN113706166A (en) * | 2021-07-21 | 2021-11-26 | 菲欧坦(重庆)数据科技有限公司 | Method for analyzing sales and after-sales satisfaction of Amazon taxonomic variants |
| CN113722496A (en) * | 2021-11-02 | 2021-11-30 | 北京世纪好未来教育科技有限公司 | Triple extraction method and device, readable storage medium and electronic equipment |
| CN117216419A (en) * | 2023-11-08 | 2023-12-12 | 江西为易科技有限公司 | Data analysis method based on AI technology |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103399916A (en) * | 2013-07-31 | 2013-11-20 | 清华大学 | Internet comment and opinion mining method and system on basis of product features |
| CN106951514A (en) * | 2017-03-17 | 2017-07-14 | 合肥工业大学 | A kind of automobile Method for Sales Forecast method for considering brand emotion |
| CN108563647A (en) * | 2017-11-29 | 2018-09-21 | 重庆邮电大学 | A kind of automobile Method for Sales Forecast method based on comment sentiment analysis |
| US20180357684A1 (en) * | 2017-01-12 | 2018-12-13 | Hefei University Of Technology | Method for identifying prefereed region of product, apparatus and storage medium thereof |
-
2020
- 2020-01-08 CN CN202010017679.0A patent/CN111242679A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103399916A (en) * | 2013-07-31 | 2013-11-20 | 清华大学 | Internet comment and opinion mining method and system on basis of product features |
| US20180357684A1 (en) * | 2017-01-12 | 2018-12-13 | Hefei University Of Technology | Method for identifying prefereed region of product, apparatus and storage medium thereof |
| CN106951514A (en) * | 2017-03-17 | 2017-07-14 | 合肥工业大学 | A kind of automobile Method for Sales Forecast method for considering brand emotion |
| CN108563647A (en) * | 2017-11-29 | 2018-09-21 | 重庆邮电大学 | A kind of automobile Method for Sales Forecast method based on comment sentiment analysis |
Non-Patent Citations (1)
| Title |
|---|
| 石文华 等: "在线初次评论与在线追加评论对商品销量影响的比较研究" * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991767A (en) * | 2019-12-20 | 2020-04-10 | 浙江大学 | Leading user identification and prediction method and technical trend prediction method |
| CN110991767B (en) * | 2019-12-20 | 2022-06-10 | 浙江大学 | Leading user identification and prediction method and technical trend prediction method |
| CN111882039A (en) * | 2020-07-28 | 2020-11-03 | 平安科技（深圳）有限公司 | Physical machine sales data prediction method and device, computer equipment and storage medium |
| WO2021139335A1 (en) * | 2020-07-28 | 2021-07-15 | 平安科技（深圳）有限公司 | Method and apparatus for predicting sales data of physical machine, and computer device and storage medium |
| CN112069402A (en) * | 2020-08-28 | 2020-12-11 | 山东省人工智能研究院 | Personalized comment recommendation method based on emotion and graph convolution neural network |
| CN113516164A (en) * | 2021-04-30 | 2021-10-19 | 辽宁工程技术大学 | Fruit tree disease and insect pest diagnosis method integrating knowledge map and deep learning |
| CN113393279A (en) * | 2021-07-08 | 2021-09-14 | 北京沃东天骏信息技术有限公司 | Order quantity estimation method and system |
| CN113706166A (en) * | 2021-07-21 | 2021-11-26 | 菲欧坦(重庆)数据科技有限公司 | Method for analyzing sales and after-sales satisfaction of Amazon taxonomic variants |
| CN113722496A (en) * | 2021-11-02 | 2021-11-30 | 北京世纪好未来教育科技有限公司 | Triple extraction method and device, readable storage medium and electronic equipment |
| CN117216419A (en) * | 2023-11-08 | 2023-12-12 | 江西为易科技有限公司 | Data analysis method based on AI technology |
| CN117216419B (en) * | 2023-11-08 | 2024-02-09 | 江西为易科技有限公司 | Data analysis method based on AI technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111242679A (en) | Sales forecasting method based on product review viewpoint mining | |
| Seerat et al. | Opinion Mining: Issues and Challenges(A survey) | |
| CN112001187A (en) | Emotion classification system based on Chinese syntax and graph convolution neural network | |
| CN112001186A (en) | Emotion classification method using graph convolution neural network and Chinese syntax | |
| Beigi et al. | Automatic construction of domain-specific sentiment lexicon for unsupervised domain adaptation and sentiment classification | |
| CN112991017A (en) | Accurate recommendation method for label system based on user comment analysis | |
| CN110706028A (en) | Commodity evaluation emotion analysis system based on attribute characteristics | |
| Wei et al. | Sentiment classification of Chinese Weibo based on extended sentiment dictionary and organisational structure of comments | |
| CN111274493B (en) | Grading prediction method based on multi-source user comments | |
| Rani et al. | Study and comparision of vectorization techniques used in text classification | |
| CN111966888B (en) | Aspect class-based interpretability recommendation method and system for fusing external data | |
| CN110110220A (en) | Merge the recommended models of social networks and user's evaluation | |
| Al-Ghuribi et al. | A comprehensive overview of recommender system and sentiment analysis | |
| CN114942974A (en) | E-commerce platform commodity user evaluation emotional tendency classification method | |
| CN117235253A (en) | Truck user implicit demand mining method based on natural language processing technology | |
| Mokadam et al. | Online product review analysis to automate the extraction of customer requirements | |
| Zhang | Application of data mining technology in the analysis of e-commerce emotional law | |
| CN113538106A (en) | Commodity refinement recommendation method based on comment integration mining | |
| CN113821718A (en) | Article information pushing method and device | |
| CN111723127A (en) | Stock trend prediction method and system based on text abstract emotion mining | |
| Pandi et al. | Reputation based online product recommendations | |
| Chebil et al. | Clustering social media data for marketing strategies: Literature review using topic modelling techniques | |
| Li et al. | The Effect of Online Reviews on E-book Pricing: A Text Analytics Approach | |
| Ye et al. | Analyzing Public Perception of Educational Books via Text Mining of Online Reviews | |
| Sun et al. | Research on ARIA of Affective Recommendation Index Algorithm Based on Topic Selection of Publications |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |