Total area of bamboo groves in Korea which is limited to
This study investigated consumer intention to use a location-based mobile shopping service (LBMSS) that integrates cognitive and affective responses. Information relevancy was integrated into pleasure-arousal-dominance (PAD) emotional state model in the present study as a conceptual framework. The results of an online survey of 335 mobile phone users in the U.S. indicated the positive effects of arousal and information relevancy on pleasure. In addition, there was a significant relationship between pleasure and intention to use a LBMSS. However, the relationship between dominance and pleasure was not statistically significant. The results of the present study provides insight to retailers and marketers as to what factors they need to consider to implement location-based mobile shopping services to improve their business performance. Extended Abstract : Location aware technology has expanded the marketer's reach by reducing space and time between a consumer's receipt of advertising and purchase, offering real-time information and coupons to consumers in purchasing situations (Dickenger and Kleijnen, 2008; Malhotra and Malhotra, 2009). LBMSS increases the relevancy of SMS marketing by linking advertisements to a user's location (Bamba and Barnes, 2007; Malhotra and Malhotra, 2009). This study investigated consumer intention to use a location-based mobile shopping service (LBMSS) that integrates cognitive and affective response. The purpose of the study was to examine the relationship among information relevancy and affective variables and their effects on intention to use LBMSS. Thus, information relevancy was integrated into pleasure-arousal-dominance (PAD) model and generated the following hypotheses. Hypothesis 1. There will be a positive influence of arousal concerning LBMSS on pleasure in regard to LBMSS. Hypothesis 2. There will be a positive influence of dominance in LBMSS on pleasure in regard to LBMSS. Hypothesis 3. There will be a positive influence of information relevancy on pleasure in regard to LBMSS. Hypothesis 4. There will be a positive influence of pleasure about LBMSS on intention to use LBMSS. E-mail invitations were sent out to a randomly selected sample of three thousand consumers who are older than 18 years old and mobile phone owners, acquired from an independent marketing research company. An online survey technique was employed utilizing Dillman's (2000) online survey method and follow-ups. A total of 335 valid responses were used for the data analysis in the present study. Before the respondents answer any of the questions, they were told to read a document describing LBMSS. The document included definitions and examples of LBMSS provided by various service providers. After that, they were exposed to a scenario describing the participant as taking a saturday shopping trip to a mall and then receiving a short message from the mall. The short message included new product information and coupons for same day use at participating stores. They then completed a questionnaire containing various questions. To assess arousal, dominance, and pleasure, we adapted and modified scales used in the previous studies in the context of location-based mobile shopping service, each of the five items from Mehrabian and Russell (1974). A total of 15 items were measured on a seven-point bipolar scale. To measure information relevancy, four items were borrowed from Mason et al. (1995). Intention to use LBMSS was captured using two items developed by Blackwell, and Miniard (1995) and one items developed by the authors. Data analyses were conducted using SPSS 19.0 and LISREL 8.72. A total of usable 335 data were obtained after deleting the incomplete responses, which results in a response rate of 11.20%. A little over half of the respondents were male (53.9%) and approximately 60% of respondents were married (57.4%). The mean age of the sample was 29.44 years with a range from 19 to 60 years. In terms of the ethnicity there were European Americans (54.5%), Hispanic American (5.3%), African-American (3.6%), and Asian American (2.9%), respectively. The respondents were highly educated; close to 62.5% of participants in the study reported holding a college degree or its equivalent and 14.5% of the participants had graduate degree. The sample represents all income categories: less than $24,999 (10.8%), $25,000-$49,999 (28.34%), $50,000-$74,999 (13.8%), and $75,000 or more (10.23%). The respondents of the study indicated that they were employed in many occupations. Responses came from all 42 states in the U.S. To identify the dimensions of research constructs, Exploratory Factor Analysis (EFA) using a varimax rotation was conducted. As indicated in table 1, these dimensions: arousal, dominance, relevancy, pleasure, and intention to use, suggested by the EFA, explained 82.29% of the total variance with factor loadings ranged from .74 to .89. As a next step, CFA was conducted to validate the dimensions that were identified from the exploratory factor analysis and to further refine the scale. Table 1 exhibits the results of measurement model analysis and revealed a chi-square of 202.13 with degree-of-freedom of 89 (p =.002), GFI of .93, AGFI = .89, CFI of .99, NFI of .98, which indicates of the evidence of a good model fit to the data (Bagozzi and Yi, 1998; Hair et al., 1998). As table 1 shows, reliability was estimated with Cronbach's alpha and composite reliability (CR) for all multi-item scales. All the values met evidence of satisfactory reliability in multi-item measure for alpha (>.91) and CR (>.80). In addition, we tested the convergent validity of the measure using average variance extracted (AVE) by following recommendations from Fornell and Larcker (1981). The AVE values for the model constructs ranged from .74 through .85, which are higher than the threshold suggested by Fornell and Larcker (1981). To examine discriminant validity of the measure, we again followed the recommendations from Fornell and Larcker (1981). The shared variances between constructs were smaller than the AVE of the research constructs and confirm discriminant validity of the measure. The causal model testing was conducted using LISREL 8.72 with a maximum-likelihood estimation method. Table 2 shows the results of the hypotheses testing. The results for the conceptual model revealed good overall fit for the proposed model. Chi-square was 342.00 (df = 92, p =.000), NFI was .97, NNFI was .97, GFI was .89, AGFI was .83, and RMSEA was .08. All paths in the proposed model received significant statistical support except H2. The paths from arousal to pleasure (H1:
Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.