Method for predicting purchase action based on time-sequence sensory evaluation

Abstract

To provide a method for predicting a product purchase action, and to provide a tempting product by manufacturing the product on the basis of the result of prediction obtained by the method.SOLUTION: The method for predicting a product purchase action causes a computer to perform the prediction step of predicting a purchase action on a single product or on a plurality of products on the basis of an answer (an experimental sampling step) obtained repeatedly on the basis of a plurality of indexes and an answer (a total evaluation step) regarding a purchase action obtained after a product is used.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for predicting product purchase behavior using time-series data collected in a natural environment in life, and a product designed using the prediction method.

Background Art

[0002] In general sensory evaluation, a small amount of a sample is taken from a product and the impression of the product is examined in a particularly controlled sensory evaluation room. On the other hand, it is considered that in daily life, consumers capture the impression while using a larger amount of the whole product in life, and these impressions do not always match.

[0003] For example, large-capacity beverages such as bottled drinks may take several hours to drink rather than finishing the whole amount in a short time. During that time, the impression of the product may change little by little. When buying a bottled product refrigerated in a store and drinking the bottled drink while doing daily activities, the temperature of the drink rises over time, and the flavor impression is likely to change. It is considered that carbonated drinks start to lose carbonation at the same time as the bottle is opened, and the carbonation feeling weakens over time. Also, there is a phenomenon called adaptation in human senses. It is known that even if the same smell or taste stimulus is received, the sensitivity and susceptibility decrease over time as the stimulus continues to be received. Not only sensory stimuli but also the emotions caused by them become accustomed over time. It is difficult to capture such changes in daily product impressions in general sensory evaluation.

[0004] The experience sampling method is a research method that elucidates the true state of people's lives by repeatedly measuring what individuals have instantaneously felt or experienced in their daily lives. By applying the experience sampling method to the sensory evaluation of an entire product, it becomes possible to capture the impressions that change while using the product and the emotional changes of consumers. For example, in Non-Patent Document 1, the experience sampling method is used to continuously evaluate food and beverages. However, no example is known in which it has been effectively applied to a method for predicting the purchasing behavior of products based on the results obtained by the experience sampling method.

Prior Art Documents

Non-Patent Documents

[0005]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] An object of the present invention is to provide a method for predicting the purchasing behavior of a product that can be applied to an attractive product design for consumers based on the sensory evaluation results utilizing the experience sampling method.

Means for Solving the Problems

[0007] The present invention has been made in consideration of the above circumstances and includes the following aspects. 〔1〕A method for predicting the purchasing behavior of a product, 1) During the use of a single or a plurality of products, Subject responses regarding at least one indicator selected from product impressions, product preferences, subject emotions, and subject behaviors A step of collecting the above two or more times (experience sampling step), and 2) After using the product, Subject responses regarding at least one indicator related to purchase behavior selected from product preferences, subject emotions, and subject behaviors A step of collecting the above (comprehensive evaluation step), and 3) Based on the responses obtained in the experience sampling step and the comprehensive evaluation step, performing a computer modeling method to predict product purchase behavior (prediction step), and The prediction method including the above. 〔2〕The prediction method according to 〔1〕, wherein step 1) is carried out in a natural environment in life and includes obtaining time series data. 〔3〕The indicator in step 1) is a) Indicators related to the product: Impression of smell, impression of taste, impression of flavor, impression of appearance, impression of temperature, impression of stimulation, impression of preference, b) Indicators related to the subject: Emotional expression, positive emotion, negative emotion, neutral emotion, previous behavior, current behavior, The prediction method according to 〔1〕 or 〔2〕, which is an indicator related to at least one selected from the above. 〔4〕The indicator related to purchase behavior in step 2) is c) Indicators related to the product: Impression of preference, overall impression, price, willingness to recommend to others, d) Indicators related to the subject: Emotional expression, positive emotion, negative emotion, neutral emotion, previous behavior, current behavior, current purchase behavior, future purchase behavior, The prediction method according to any one of 〔1〕 to 〔3〕, which is an indicator related to at least one selected from the above. 〔5〕The modeling method in step 3) is Simple regression, multiple regression, logistic regression, Ridge regression, Lasso regression, elastic net regression, PLS regression, support vector machine (SVM), decision tree, regression tree, random forest, gradient boosting, naive Bayes, neural network, convolutional neural network (CNN), recurrent neural network (RNN), graph neural network (GNN), and multilevel analysis, time series analysis, The prediction method according to any one of [1] to [4], which is at least one selected from . [6] The prediction method according to any one of [1] to [5], wherein the product is a food or beverage or a cosmetic. [7] A fragrance composition, a food or beverage, or a cosmetic designed by using the prediction method according to any one of [1] to [6].

Advantages of the Invention

[0008] According to the present invention, in the quality evaluation of products, it is possible to provide a prediction method for product making that consumers actually want to buy in their lives, which is difficult with conventional sensory evaluation and preference survey methods. By using this method, it is possible to efficiently design products that consumers want to buy.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Modes for Carrying Out the Invention

[0010] Hereinafter, embodiments of the present disclosure will be described in detail. The present disclosure is not limited to the following embodiments and can be implemented with modifications within the scope of the gist of the invention.

[0011] <Method for Predicting Product Purchase Behavior> The method for predicting the purchasing behavior of the product of the present invention includes, as essential steps, an experience sampling step, an overall evaluation step, and a prediction step, which will be described later. Further, as long as the effects of the present invention are not impaired, other steps may be further included.

[0012] <Experience Sampling Step> (Indicators in the Experience Sampling Step) As the indicators in the experience sampling step of the present invention, they can be arbitrarily selected according to the purpose of evaluation, but indicators that change due to continuous use of the product are preferred. Specifically, it is preferably at least one indicator selected from the impression of the product, the preference for the product, the emotions of the subject, and the behavior of the subject. In particular, by answering the impression and preference of the new product that is fresh in memory within 2 hours, the emotions and behavior of the subject, an answer without distortion due to memory can be obtained.

[0013] When taking indicators related to the product, they can be arbitrarily selected according to the purpose of evaluation, and examples include the impression of fragrance, the impression of taste, the impression of flavor, the impression of appearance, the impression of temperature, the impression of stimulation, the impression of preference, etc. For example, it includes terms for describing sensations such as flavor descriptors, taste sensations such as sweetness, bitterness, saltiness, sourness, umami and astringency, olfactory sensations such as sweet fragrance, grassy smell, acid smell, roasted fragrance, floral fragrance, etc., described in Lexicon for Sensory Evaluation: Aroma, Flavor, Texture and Appearance (G.V. Civille et al., ASTM International, West Conshohocken, PA, 2011), physical sensations such as spiciness, stimulation, warm feeling, cold feeling, firm, smooth, etc., and auditory sensations such as chewing sound, swallowing sound, etc. Furthermore, as the types of terms for describing the above sensations, onomatopoeias unique to Japanese such as crispy, sticky, crispy, crackling, swishy, rustling, fluffy, crunchy, squishy, slow-flowing, refreshing, chilly, stinging, prickly are also included. Also, as the sensory characteristics indicating the preference for flavor, deliciousness, good balance, preference, and whether one wants to eat it are also included. Furthermore, as the types of terms for describing appearance, color, brightness, size, preference, etc. are also included.

[0014] When using it as an index related to the subject, it can be arbitrarily selected according to the purpose of evaluation. Examples include emotional expressions, positive emotions, negative emotions, neutral emotions, immediate actions, current actions, etc. For example, emotional expressions include both verbal and non-verbal expressions, and can be arbitrarily selected according to the purpose of evaluation. For example, emotional impressions such as beautiful, cute, interesting, warm, and gorgeous described in the study of emotional expression terms (1) (Kurosuke et al., 77th Annual Convention of the Japanese Psychological Association, 2013); emotional judgments such as clear, ambiguous, high-class, non-high-class, modern, old-fashioned, innovative, conservative, masculine, feminine, foreign-style, Japanese-style; emotional expressions such as good, bad, like, dislike, beautiful, ugly, comfortable, uncomfortable, good at, bad at, easy, difficult, etc.

[0015] Emotions include positive emotions, negative emotions, and neutral emotions, and can be arbitrarily selected according to the purpose of evaluation. For example, the six basic emotions of joy, surprise, anger, fear, disgust, and sadness proposed by Ekman; the eight basic emotions of joy, anger, sadness, surprise, horror, disgust, acceptance, and expectation proposed by Plutchik; the attention, excitement, vitality, happiness, satisfaction, brightness, relaxation, calmness, fatigue, listlessness, depression, sadness, worry, stress, nervousness, etc. consisting of the dimensions of arousal and pleasure-displeasure proposed by Russell. Furthermore, it also includes more complex emotions such as familiarity, nostalgia, relaxation, refreshment, excitement, concentration, stress, boredom, flow, creativity, mindfulness, mind wandering, and well-being.

[0016] Actions include the previous action and the current action, and can be arbitrarily selected according to the purpose of evaluation. For example, specific actions include sleep, which is a primary activity necessary for physiology, daily chores, meals, commuting and going to school, which are secondary activities with a strong obligatory nature in social life, work, study, housework, caregiving and nursing, childcare, shopping, movement in tertiary activities during free time, TV, radio, newspapers, magazines, rest and relaxation, learning and research, hobbies and entertainment, sports, social activities, socializing and dating, medical treatment and recuperation, etc., and activities described in the types of actions in the Basic Survey of Social Life by the Statistics Bureau of the Ministry of Internal Affairs and Communications. Furthermore, it also includes actions related to products such as eating behavior including the progress, remaining amount, appetite, etc. of eating, and usage behavior such as the remaining amount of cosmetics. It also includes not only conscious actions but also unconscious actions.

[0017] The number of indicators in the experience sampling step can be arbitrarily selected as one or more according to the purpose of evaluation. The answers to the indicators in the experience sampling step include quantification by nominal scale, ordinal scale, sensory scale, ratio scale, and free description, and can be arbitrarily selected according to the purpose of evaluation. It is preferable to implement multiple answers by the same method. For example, the number of selections and the number of people who select by Check-All-That-Apply (CATA), a rating scale of 2 to 9 items, continuous values by Visual Analogue Scale (VAS), Labeled Magnitude Scale (LMS), frequency, probability, vector, co-occurrence number, etc. by free description are included.

[0018] (Timing of answers in the experience sampling step) In the present invention, as the timing of responses in the experience sampling step, there are multiple points in time two or more times from the start to the end of product use, and the number of times and timing can be arbitrarily selected according to the purpose of evaluation. The multiple points in time may be the same or different among the subjects. Furthermore, the multiple points in time may be event-based or time-based, and may be fixed in advance or randomized. For example, as responses based on events, there are points in time when a certain amount of the product has been consumed, when a break has been taken, when housework has been done, etc. As responses based on time, there are points in time when a certain period of time has elapsed, certain times, time zones such as morning, noon, and night, etc.

[0019] (Method for collecting responses in the experience sampling step) In the present invention, as the method for collecting responses in the experience sampling step, either paper-based or web-based can be arbitrarily selected according to the purpose. It may be the same or different for the subjects, but the same is more efficient for collecting responses. When collecting responses in a paper-based manner, a format such as a self-descriptive diary consisting of multiple sheets is preferred. When collecting responses in a web-based manner, access may be made by distributing a web link in advance, but by distributing it immediately before, the subject can answer without forgetting.

[0020] (Natural environment in life) The experience sampling step of the present invention is preferably carried out in a "natural environment in life". The "natural environment in life" is not particularly limited, but includes the environment in which the product is actually used. The environment in which the product is used is appropriately selected according to the type and use of the product, etc. It is preferably not in a particularly controlled sensory evaluation room or psychological experiment room.

[0021] When the product is a food or beverage, as the "natural environment in life", the natural environment for consuming the food or beverage can be arbitrarily selected from time of day, duration of eating or drinking, temperature, season, location, activity, etc. Examples of activities include primary activities that are physiologically necessary such as daily chores, eating, secondary activities with a strong obligatory nature in conducting social life such as commuting to work or school, working, studying, housework, caregiving, childcare, shopping, tertiary activities during free time such as moving around, watching TV / listening to the radio / read newspapers / magazines, resting / relaxing, learning / researching, hobbies / entertainment, sports, social activities, socializing, visiting a doctor / undergoing medical treatment, etc., and activities described in the types of actions in the Basic Survey on Social Life by the Statistics Bureau of the Ministry of Internal Affairs and Communications.

[0022] When the product is a cosmetic, as the "natural environment in life", the natural environment for using the cosmetic can be arbitrarily selected from time of day, duration of use, temperature, season, location, activity, etc. Examples of activities include primary activities that are physiologically necessary such as sleeping, daily chores, eating, secondary activities with a strong obligatory nature in conducting social life such as commuting to work or school, working, studying, housework, caregiving, childcare, shopping, tertiary activities during free time such as moving around, watching TV / listening to the radio / read newspapers / magazines, resting / relaxing, learning / researching, hobbies / entertainment, sports, social activities, socializing, visiting a doctor / undergoing medical treatment, etc., and activities described in the types of actions in the Basic Survey on Social Life by the Statistics Bureau of the Ministry of Internal Affairs and Communications.

[0023] <Comprehensive Evaluation Step> (Indicators in the Comprehensive Evaluation Step) As the indicators in the comprehensive evaluation step of the present invention, they can be arbitrarily selected according to the purpose of evaluation, but indicators related to purchasing behavior are preferred. Specifically, at least one selected from product preference, the emotions of the subject, and the behavior of the subject is preferred.

[0024] When taking indicators related to the product, they can be arbitrarily selected according to the purpose of evaluation, and examples include the impression of preference, the overall impression, price, etc. For example, as the impression of preference, there are deliciousness, pleasure or discomfort, like or dislike; as the overall impression, there are richness, satisfaction, something to recommend to others; as the price, there are expected price, Willingness To Pay (WTP), etc.

[0025] When using an index related to the subject, it can be arbitrarily selected according to the purpose of evaluation. However, an index related to purchasing behavior is preferred from among emotional expressions, positive emotions, negative emotions, neutral emotions, current behavior, future behavior, etc. For example, emotional expressions include both verbal and non-verbal expressions, and can be arbitrarily selected according to the purpose of evaluation. For example, emotional impressions such as beautiful, cute, interesting, warm, and gorgeous, emotional judgments such as clear, ambiguous, high-class, non-high-class, modern, old-fashioned, innovative, conservative, masculine, feminine, foreign-style, and Japanese-style described in the study of emotional expression terms (1) (Kurosuzu et al., 77th Annual Convention of the Japanese Psychological Association, 2013), and emotional expressions such as good, bad, like, dislike, beautiful, ugly, comfortable, uncomfortable, good at, bad at, easy, and difficult can be cited.

[0026] Emotions include positive emotions, negative emotions, and neutral emotions, and can be arbitrarily selected according to the purpose of evaluation. For example, the six basic emotions of joy, surprise, anger, fear, disgust, and sadness proposed by Ekman, the eight basic emotions of joy, anger, sadness, surprise, horror, disgust, acceptance, and expectation proposed by Plutchik, and attention, excitement, vitality, happiness, satisfaction, brightness, relaxation, calmness, fatigue, listlessness, depression, sadness, worry, stress, nervousness, and tension consisting of the dimensions of arousal and pleasure-displeasure proposed by Russell can be cited. Furthermore, more complex emotions such as familiarity, nostalgia, relaxation, refreshment, concentration, stress, boredom, flow, creativity, mindfulness, mind wandering, and well-being are also included.

[0027] Behavior includes both current behavior and future behavior, and can be arbitrarily selected according to the purpose of evaluation. In particular, as an index related to purchasing behavior, purchase intention, purchase selection, the price actually paid, etc. are preferred. Furthermore, behavior related to products such as the progress and remaining amount of eating, and the remaining amount of cosmetics used is also included. Also included is not only conscious behavior but also unconscious behavior.

[0028] The number of indicators in the comprehensive evaluation step can be arbitrarily selected from one or more types according to the purpose of evaluation.

[0029] The responses to the indicators in the comprehensive evaluation step include quantification by nominal scale, ordinal scale, sensory scale, ratio scale, and free description, and can be arbitrarily selected according to the purpose of evaluation. It is desirable to perform multiple responses by the same method. For example, the number of selections and the number of people making selections by CATA, a rating scale of 2 to 9 items, continuous values by Visual Analogue Scale (VAS), Labeled Magnitude Scale (LMS), frequency, probability, vector, co-occurrence number, etc. by free description can be mentioned.

[0030] (Timing of responses in the comprehensive evaluation step) In the present invention, as the timing of responses in the comprehensive evaluation step, it is after the experience sampling step is completed, but the period can be arbitrarily selected according to the purpose. The timing of the comprehensive evaluation may be the same or different among the subjects. Furthermore, the timing of the comprehensive evaluation may be event-based or time-based, and may be fixed in advance or randomized. For example, as event-based responses, the time when one wants to eat again, the time when a purchase behavior occurs, etc. can be mentioned. As time-based responses, the time when a certain period of time has elapsed, a certain time, time zones such as morning, noon, and night can be mentioned.

[0031] (Method for collecting responses in the comprehensive evaluation step) In the present invention, as the method for collecting responses in the experience sampling step, either paper-based or web-based can be arbitrarily selected according to the purpose. It may be the same or different among the subjects, but the same is more efficient for collecting responses. Responses can be efficiently collected by using the same method as the experience sampling step, but it is preferable that the responses in the experience sampling step cannot be referred to.

[0032] <Prediction step> (Definition of the prediction step) In the present invention, the prediction step refers to a step of predicting the purchasing behavior of a product by performing a computer modeling method based on the responses obtained in the empirical sampling step and the comprehensive evaluation step. For example, it includes a model with the purchase intention as the target variable and the deliciousness of the product as the explanatory variable, a model with the deliciousness related to the purchasing behavior as the target variable and the impression of the product as the explanatory variable, a model with the emotion related to the purchasing behavior as the target variable and the impression of the product as the explanatory variable, and the like.

[0033] (Modeling method in the prediction step) The modeling method in the prediction step of the present invention can be arbitrarily selected according to the purpose, and includes simple regression, multiple regression, logistic regression, Ridge regression, Lasso regression, elastic net regression, PLS regression, support vector machine (SVM), decision tree, regression tree, random forest, gradient boosting, naive Bayes, neural network, convolutional neural network (CNN), recurrent neural network (RNN), graph neural network (GNN), and multilevel analysis, time series analysis. These modeling methods may be used alone or in combination.

[0034] (Explanatory variable in the prediction step) The explanatory variable in the prediction step of the present invention can be arbitrarily selected according to the purpose, but a numerical value based on the response collected in the empirical sampling step is preferred. Also, the number of explanatory variables can be arbitrarily selected as long as it is one or more, but two or more for capturing time series changes are preferred. For example, either one or both of the indicators related to the product and the indicators related to the subject may be used. It also includes values obtained by adding, subtracting, taking ratios of time series changes, change patterns, smoothed patterns, AUC, and the like.

[0035] (Target variable in the prediction step) The target variable in the prediction step of the present invention can be arbitrarily selected according to the purpose, but a numerical value or label related to purchasing behavior based on the responses collected in the comprehensive evaluation step is preferable. Also, the number of target variables can be arbitrarily selected as long as it is one or more. For example, either one or both of the indicators related to the product and the indicators related to the subject may be used. It also includes values obtained by adding, subtracting, taking ratios, or weighted adding the responses of multiple types of indicators.

[0036] (Selection criteria for the model in the prediction step) The selection criteria for the model in the prediction step of the present invention can be arbitrarily selected according to the purpose, and include, for example, R, RMSE, MAE, coefficient of determination R-squared, AIC, BIC, prediction accuracy, readability, goodness of fit, correct answer rate, and reproducibility.

[0037] <Product> In the present invention, the products to be evaluated are not particularly limited, but food and beverages and cosmetics are particularly preferred.

[0038] Examples of foods include dairy products, confectioneries, oil products, agricultural processed products, seasonings, soups, livestock processed products, fishery processed products, and other foods.

[0039] Examples of dairy products include butter, cheese, and cheese food.

[0040] Examples of confectioneries include frozen desserts, Japanese confectioneries, Western confectioneries, desserts, baked confectioneries, bakeries, and candies.

[0041] Examples of oil products include margarine, coffee whiteners, and chocolates.

[0042] Examples of agricultural processed products include noodles, vegetable protein processed products, jams, pastes, pickles, canned agricultural products, fruit juices, and fruit pulp processed products.

[0043] Examples of seasonings include miso, soy sauce, sauces, mayonnaise, and dressings.

[0044] Examples of soups include powdered soups, retort pouch soups, and canned soups. Examples of livestock processed products include hams, sausages, hamburgers, and canned meat.

[0045] Examples of fishery processed products include fish hams, fish sausages, fishery paste products, and canned fishery products.

[0046] Examples of other foods include bakery products, candies, gums, frozen foods, retort foods, instant foods, livestock feeds, fish farming feeds, pet feeds, oral care products, and the like.

[0047] In addition, raw materials such as milk, grains, edible oils, fruits, agricultural products such as vegetables, livestock, and seafood are also targeted.

[0048] Examples of beverages include carbonated beverages, fruit beverages, vegetable beverages, beers, non-alcoholic beers, non-alcoholic beverages, chu-hais, other alcoholic beverages, tea beverages, coffee beverages, functional beverages, cocoa beverages, cacao beverages, sugar-free beverages, sports beverages, nutritional and tonic drinks, dairy products, lactic acid bacteria beverages, milk beverages, vinegar beverages, and the like.

[0049] Examples of teas include tea leaves and tea extracts that are raw materials for green tea, black tea, oolong tea, and the like.

[0050] Coffee beverages include coffee beans, coffee liquids, and instant coffee.

[0051] Preferred beverages include coffee beverages and tea beverages, and particularly preferred is coffee beverage.

[0052] The form of the beverage product is not particularly limited, and examples include container-packed beverages (RTD (Ready-To-Drink) beverages) such as cans and PET bottles that can be drunk as they are after purchase, powdered beverages that are dissolved in water or hot water and then consumed, and products such as tea and coffee-based beverages that require leaching treatment for drinking.

[0053] Food and drink products also include flavor raw materials.

[0054] Examples of flavor raw materials include citrus such as lemon, grapefruit, and orange; fruits such as apple, melon, grape, peach, and pineapple; favorite beverages such as black tea, green tea, oolong tea, and coffee; dairy products such as milk and yogurt; vanilla; mint such as peppermint, spearmint, and Japanese mint; spices such as black pepper, cinnamon, clove, nutmeg, turmeric, basil, oregano, rosemary, sage, thyme, cardamom, coriander, dill, and fennel; nuts such as almond, cashew nut, peanut, walnut, and pine nut; meats such as beef, pork, and chicken; and seafood raw materials such as salmon, skipjack tuna, wakame seaweed, kelp, crab, and scallop. If it is an origin substance of plant origin, these fruits, flowers, buds, trees, barks, branches, leaves, stems, roots, and their extracts, etc. can be targeted.

[0055] Examples of flavor raw materials other than these include the origin substances described in the "Collection of Natural Flavor Raw Materials" (published by the Japan Flavor Industry Association on June 1, 2011).

[0056] Food and drink products also include food additives.

[0057] Examples of food additives include, for example, fragrances (synthetic fragrances, natural fragrances, compound fragrances), sweeteners (acesulfame potassium, stevia, erythritol, etc.), acidulants (citric acid, tartaric acid, phosphoric acid, lactic acid, etc.), bittering agents (caffeine, naringin, etc.), seasonings (sodium glutamate, L-arginine, etc.), colorants (chlorophyllin, grape skin pigment, safflower red pigment, Food Red No. 102, sodium copper chlorophyllin, etc.), preservatives (benzoic acid, sorbic acid, etc.), thickening and stabilizing agents (carrageenan, xanthan gum, guar gum, dextran, pullulan, etc.), emulsifiers (chicle extract, enzymatically decomposed lecithin, glycerin fatty acid ester, sucrose fatty acid ester, polysorbate 60, etc.), manufacturing agents (sodium hydrogen carbonate, potassium carbonate, magnesium carbonate, etc.), antioxidants (catechin, quercetin, tea extract, raw coffee bean extract, d-α-tocopherol, calcium disodium ethylenediaminetetraacetate, etc.), color formers (potassium nitrate, sodium nitrate, etc.), brighteners (carnauba wax, lanolin, paraffin wax, etc.), bleaching agents (calcium acetate, sodium metabisulfite, etc.), enzymes (lipase, pectinase, polyphenol oxidase, protease, etc.), and other additives that can be used in food and beverages, excipients (dextrin, gum arabic, corn starch, modified starch, etc.), spices (mint, pepper, perilla, garlic, ginger, etc.), inorganic salts (sodium chloride, potassium chloride, etc.), and other flavor improvers (flavor-improving peptides, juice-derived fractions, etc.).

[0058] The cosmetics in the present invention are not particularly limited, and examples include fragrance products, basic cosmetics, finishing cosmetics, hair cosmetics, suntan cosmetics, pharmaceutical cosmetics, hair care products, soaps, body cleansers, bath agents, detergents, softeners, cleaning agents, kitchen detergents, bleaching agents, aerosol agents, consumer goods, air fresheners, repellents, toothpastes, oral care products, etc.

[0059] Among these, shampoo, laundry detergent, bath salts, dishwashing detergent, hair products (hair styling aids, styling agents, etc.), cosmetics (body wash, body lotion, cologne, powder foundation, etc.), bathroom cleaners, floor cleaners, air fresheners, deodorants (rug deodorants, room deodorants, etc.), and window cleaners are particularly preferred.

[0060] The number of products to be subjected to sensory evaluation can be arbitrarily selected depending on the purpose of the evaluation. However, particularly when aiming at product development, evaluating and comparing two or more products can more explicitly express the flavor characteristics of complex food and beverage products. When evaluating two or more products for the purpose of product development, examples of combinations of food and beverage products include differences in raw materials such as differences in the origin of raw materials and differences in raw material lots, differences in the types and amounts of food additives such as flavor compositions, emulsifiers, and pH adjusters, the presence or absence and order of processing steps such as heating steps, distillation steps, concentration steps, and filtration steps, differences in parameters of each step, controllable factors such as differences in the temperature and amount of the products subjected to sensory evaluation, and labeling factors and error factors such as the state of crystals, the distribution of emulsion particles, and the moisture content.

Example

[0061] Hereinafter, the present invention will be described more specifically with reference to examples, but the present invention is not limited to these examples in any way.

[0062] (Example 1) (Sensory Evaluation and Prediction of Purchase Intention of Latte-Type PET Bottle Coffee) Method

[0063] <Experience Sampling Step> (Sample) Four labels of latte-type (type with added milk) bottled coffee products were removed, placed in a plastic bottle holder, and blinded. A seal with the sample identification numbers (A, B, C, D) was attached to the cap. The plastic bottle holder was cut vertically at one place to allow the remaining amount of the sample to be visible. The plastic bottles were pre-marked every 125 ml of the content volume, and a total of five marks were made from the first sip to the end of drinking (see Fig. 2 for details). They were stored in the refrigerator until just before the evaluation. (Subjects (hereinafter also referred to as "panels").) Twenty-three in-house preference panels participated in the evaluation. (Procedure) The panels participated in the evaluation for four days, and evaluated one type of sample per day. The evaluation order of the samples was counterbalanced to control among the panels. The panels received one sample at 9:00 am and drank the one sample per day at their preferred timing and in their preferred amount while performing normal work at their desks in the company. When they reached the mark on the plastic bottle, they answered questions about the current sample's flavor impression and the panel's feelings from the web link distributed in advance. They answered the same questionnaire items each time when they reached each of the five marks. At that time, the items regarding the flavor impression of the sample were subjective coldness, overall aroma intensity, sweet aroma, flowery aroma, smoky, roasted feeling, chocolate-like flavor, vanilla-like flavor, sweetness, bitterness, sourness, clarity, off-flavors and odors, richness, thirst-quenching feeling, artificial, coffee feeling, milk feeling, deliciousness, and desire to continue drinking, a total of 20 items, and all were evaluated using a VAS (Visual Analogue Scale) indicating intensity from 0 to 100. The items indicating the panel's own feelings were six items: stress of the previous task, wakefulness, relaxation, refreshment, concentration, and happiness, and were also evaluated using a VAS indicating intensity. They selected the previous action from the options (experiment, desk work, meeting / coordination, movement, break, others), and at the same time, free comments regarding the current action were also collected.

[0064] <Comprehensive Evaluation Step> (Panel) Same as the above eating and drinking evaluation. (Procedure) After the eating evaluation, an evaluation was conducted on the overall impression of the sample. Regarding the items related to the impression of flavor at that time, the two items were deliciousness and the desire to keep drinking, and both were evaluated using a VAS indicating an intensity of 0 to 100. Regarding the items related to purchasing behavior, the two items were purchase intention and WTP (Willingness To Pay). The purchase intention was evaluated using a VAS indicating an intensity of 0 to 100, and for WTP, values of 0 yen or more were answered in 1 yen increments.

[0065] <Prediction step> A multiple regression model was created to predict the purchase intention from the deliciousness at five time points by setting the objective variable as the intensity of the purchase intention and the explanatory variables as the intensity of deliciousness at each time point. When variable selection was performed using the stepwise method, a multiple regression equation was obtained with the deliciousness of the first sip and the deliciousness at the 500 ml point (end of drinking) as the explanatory variables (see Fig. 3 for reference). R, which represents the prediction accuracy of the created model, was as high as 0.855. R (version 4.2.3) was used in the above prediction step.

[0066] (Comparative Example 1) (Sensory evaluation of latte-type bottled coffee and prediction of purchase intention)

[0067] <Prediction step> Based on the responses obtained by the same method as in Example 1 above, a multiple regression model was created to predict the purchase intention from only the deliciousness of the first sip by setting the objective variable as the intensity of the purchase intention and the explanatory variable as the intensity of the deliciousness of the first sip. R, which represents the prediction accuracy of the created model, was 0.684.

[0068] It was found that the prediction accuracy of the model created in Example 1 was higher than that of the model created in Comparative Example 1, and purchase behavior could be predicted better.

[0069] (Example 2) (Sensory evaluation of black-type bottled coffee and prediction of purchase intention) Four black-type bottled coffee products were used as samples.

[0070] The evaluation items regarding the flavor impression of the samples were 18 items, namely subjective coldness, overall aroma intensity, sweet aroma, flowery aroma, smoky, roasted, flavors such as chocolate, flavors like roasted sweet potato, flavors like barley tea, bitterness, sourness, off-flavors, richness, artificial, thirst-quenching, coffee flavor, deliciousness, and the desire to keep drinking.

[0071] For the others, an evaluation and prediction model was created using the same method as in Example 1 above. As a result, the R representing the prediction accuracy was 0.861.

[0072] (Comparative Example 2) (Sensory Evaluation and Prediction of Purchase Intention of Black-Type PET Bottle Coffee) Based on the responses obtained by the same method as in Example 1 above, with the objective variable being the intensity of the purchase intention and the explanatory variable being the intensity of the deliciousness of the first sip, a multiple regression model for predicting the purchase intention from only the deliciousness of the first sip was created. The R representing the prediction accuracy of the created model was 0.691.

[0073] It was found that the prediction accuracy of the model created in Example 2 was higher than that of the model created in Comparative Example 2, and the purchase behavior could be predicted better.

[0074] (Example 3) (Sensory Evaluation and Prediction of Purchase Intention of Low-Sugar-Type PET Bottle Coffee) Three low-sugar-type PET bottle coffee products were used as samples.

[0075] The evaluation items regarding the flavor impression of the samples were 17 items, namely subjective coldness, overall aroma intensity, sweet aroma, flowery aroma, roasted feeling, flavors such as chocolate, sweetness, bitterness, sourness, clarity, richness, thirst-quenching, artificial, coffee flavor, milk flavor, deliciousness, and the desire to keep drinking.

[0076] For the others, an evaluation and prediction model was created using the same method as in Example 1 above. As a result, the R representing the prediction accuracy was 0.766.

[0077] (Comparative Example 3) (Sensory Evaluation and Prediction of Purchase Intention of Low-Sugar Type PET Bottle Coffee) Based on the responses obtained by the same method as in Example 3 above, with the objective variable being the intensity of purchase intention and the explanatory variable being the intensity of deliciousness of the first sip, a multiple regression model for predicting the purchase intention from only the deliciousness of the first sip was created. The R representing the prediction accuracy of the created model was 0.222.

[0078] It was found that the prediction accuracy of the model created in Example 3 was higher than that of the model created in Comparative Example 3, and the purchase behavior could be predicted better.

[0079] (Example 4) (Sensory Evaluation and Prediction of Purchase Intention of Sugar-Free Carbonated Beverages) Two PET bottle-packed sugar-free carbonated water products were used as samples.

[0080] The evaluation items regarding the flavor of the samples were eight items: carbonic acid stimulation, throat flushing stimulation, bubble size, bitterness, carbonic acid pleasantness / unpleasantness, subjective coldness, quenching thirst feeling, and deliciousness. The items regarding the panelists' own feelings were four items: pleasantness / unpleasantness, arousal / sedation, concentration, and refreshment feeling. As a result of creating an evaluation and prediction model by the same method as in Example 1 above, the coefficient of determination R-squared representing the prediction accuracy was 0.605.

[0081] (Comparative Example 4) (Sensory Evaluation and Prediction of Purchase Intention of Sugar-Free Carbonated Beverages) Based on the responses obtained by the same method as in Example 4 above, with the objective variable being the intensity of purchase intention and the explanatory variable being the intensity of deliciousness of the first sip, a multiple regression model for predicting the purchase intention from only the deliciousness of the first sip was created. The coefficient of determination R-squared representing the prediction accuracy of the created model was 0.553.

[0082] It was found that the prediction accuracy of the model created in Example 4 was higher than that of the model created in Comparative Example 4, and the purchase behavior could be predicted better.

Claims

1. A method for predicting the purchasing behavior of products, comprising: 1) During the use of a single or multiple products, collecting the responses of the subject regarding at least one indicator selected from the impression of the product, the preference for the product, the emotions of the subject, and the behavior of the subject two or more times (experience sampling step); 2) After the use of the product, collecting the responses of the subject regarding at least one indicator related to the purchasing behavior selected from the preference for the product, the emotions of the subject, and the behavior of the subject (comprehensive evaluation step); 3) Based on the responses obtained in the experience sampling step and the comprehensive evaluation step, performing a computer-based modeling method to predict the purchasing behavior of the product (prediction step). The prediction method as described above.

2. The prediction method according to Claim 1, wherein step 1) is carried out in a natural environment in life and includes obtaining time-series data.

3. The indicator in step 1) is a) Indicators related to the product: impression of smell, impression of taste, impression of flavor, impression of appearance, impression of temperature, impression of stimulation, impression of preference, b) Indicators related to the subject: emotional expression, positive emotion, negative emotion, neutral emotion, previous behavior, current behavior, and is an indicator related to at least one selected from the above. The prediction method according to Claim 1 or 2.

4. The indicator related to the purchasing behavior in step 2) is c) Indicators related to the product: impression of preference, overall impression, price, recommend to others, d) Indicators related to the subject: emotional expression, positive emotion, negative emotion, neutral emotion, previous behavior, current behavior, current purchasing behavior, future purchasing behavior, and is an indicator related to at least one selected from the above. The prediction method according to any one of Claims 1 to 3.

5. The modeling method in step 3) is simple regression, multiple regression, logistic regression, Ridge regression, Lasso regression, elastic net regression, PLS regression, support vector machine (SVM), decision tree, regression tree, random forest, gradient boosting, naive Bayes, neural network, convolutional neural network (CNN), recurrent neural network (RNN), graph neural network (GNN) and multilevel analysis, time series analysis, and is at least one selected from the above. The prediction method according to any one of Claims 1 to 4.

6. ​ ​ The prediction method according to any one of claims 1 to 5, wherein the product is a food, drink, or cosmetic.

7. A fragrance composition, food, drink, or cosmetic designed using the prediction method according to any one of claims 1 to 6.

Images