1.1 RESEARCH BACKGROUND
In Malaysia, it's easier than ever to launch a profitable home-based or internet business. Furthermore, the expansion of home enterprises is fuelled by an increase in demand for internet shopping and freelancing professionals. Hobbies and skillset are the only constraints on the type of business to be establish. Especially women’s who are a single mother, house wife who are mostly at home who are looking for an income but unsure what to sell or how to ensure their business is growing. Home is a good place to start taking the first step. (Incifile,2021) Fundera claims that more women are starting their own enterprises. The US has 12.3 million female-owned enterprises. The number of female business owners and entrepreneurs has increased by 114% in the last two decades. Ahead of the 2020 pandemic, American Express reported that every day in 2019 roughly 1,817 new women-owned businesses were launched. Why is this stat vital? As normalcy returns, should expect to see a growth in small and home-based enterprises. Now is the moment to seize the opportunity and establish own business. This shows that home-based enterprises might be a useful answer for women entrepreneurs who lack access to beginning funding or who are juggling employment, family duties, and childcare. Women frequently seek greater independence and flexibility when starting a business, and working from home allows them to have both - a good salary and a flexible schedule. With the help of sentiment analysis, it can help home-based business growth tremendously by giving sellers insights from customers feedback. They will be able to improve product quality, socially engaged with customers and help in growth of business.
1.2 PROBLEM STATEMENT
Customer feedback is crucial for every business. It aids in product development, marketing, and operations, and in among other things. However, it is often neglected as one of the most critical components of running a business. Sentiment analysis, which recognises the emotions and feelings expressed in user-generated online content, has been presented as a solution to deal with and fix existing rating difficulties. The problem in Malaysia, as Malaysian are so used using “Bahasa rojak” or to understand better mix few languages. Most customer feedback in e-commerce website shows how mixed language, short forms are being used almost in every customer feedbacks. In this research our aim is to take in considerate of the customer feedback with employing lexicon in Bahasa Malaysia and then implementing sentiment analysis on the processed and cleaned feedback to determine whether rating systems accurately reflect true customer experience, satisfaction, and thoughts.
1.3 RESEARCH QUESTION
How is lexicon implement in customer feedback for home-based product.
What is the sentiment feature customer feedback for home-based product.
To support this research question, I establish the objective of this research is to examine Shopee customers feedback on home-based product, as following:
To implement lexicon on customer feedback on home-based product.
To determine the sentiment analysis of customers feedback on home-based product.
2.1 How E-Commerce platform has taken over offline purchases.
Once upon a time, the online and offline worlds were diametrically opposed. However, with the rise of e-commerce, particularly mobile e-commerce, the line between online and offline is rapidly blurring. As aware of the current pandemic, most of the countries in the world went into major lockdown in March 2020, forcing many businesses to temporarily close. Countries are gradually loosening their restrictions, but the future remains uncertain. Even reopened businesses have restrictions enforcing social distancing, mask wearing, and limits on how many customers can enter a space at one time. People are increasingly inclined to shop online when traditional shopping becomes difficult, if not frightening. The fact that consumers were already embracing e-commerce websites made the transition much easier. Speaking specifically on e-commerce in Malaysia, according to (Global data, 2020) comments that Malaysia is one of Southeast Asia's fastest-growing e-commerce markets, owing to rising Internet and smartphone penetration, a growing middle-class population, and tech-savvy millennials. The fear of virus transmission through cash handling and visits to physical stores as a result of the COVID-19 outbreak has accelerated this growth. Also quoted by (International Trade Administration), because of its high internet and mobile connectivity, Malaysia has a high rate of eCommerce usage. Approximately half of the population (16.29 million) is an active online shopper, and 82.9 percent of mobile users shop online. Malaysians are driven primarily by quality of product, price advantages, product variety, and the availability of reviews. Exclusive deals, free shipping, convenience, and offers from online stores all influence shoppers.
2.2 Online Shopping Platforms in Malaysia
In Malaysia, shopping is a national pastime. Malaysians are window shopping online while Malaysians seek refuge from the tropical heat in air-conditioned shopping malls. However, not all online shopping sites are created equal. Some are better than others for specific categories of items, or they may have rewards programmes that can help you save money on future purchases, or they may be better for obscure items that are difficult to find elsewhere. There are few popular online shopping sites that are very well known and is used by most of the Malaysians. Here is top nine most popular e-commerce website according to (imoney, 2021) Lazada, as of January 2020, Lazada was the most searched shopping platform in Malaysia. It was founded in 2012 and was later acquired by Chinese e-commerce behemoth Alibaba. Next is Shopee, Shopee is Malaysia's most popular e-commerce platform, and it even tops the list in Southeast Asia. This platform was only launched in 2015, but it has quickly gained a reputation for being slightly less expensive than other platforms. Besides that, Mudah.my is a well-known Malaysian online classifieds marketplace that was founded in 2007. It was later purchased by Carousell, another online marketplace platform. Followed by the rest of the e-commerce platform such as Taobao, Harvey Norman, Signature Market, Carousell, eBay, Amazon and etc. In this research project, will be focusing mainly on Shopee platform only.
Shopee is an online shopping platform that provides customers with a simple, secure, fast, and enjoyable online shopping experience. Shopee offers a diverse product selection, backed up by integrated payments and seamless fulfilment, to make it easier for both buyers and sellers. With these benefits, Shopee now has tens of millions of users on a daily basis. Shopee is also committed to assisting brands and sellers in their e-commerce success and is highly tailored to each market in which it operates. From (New straights time, 2021) quoted that with the number of sellers more than doubling in the last year, Shopee continues to see more Malaysians pivoting or starting their businesses online. This demonstrates the growing involvement of micro, small, and medium-sized enterprises (MSMEs) in the digital economy as consumers' reliance on e-commerce and digital payments grows. In addition, Shopee reported that 80 local sellers who joined the e-commerce platform in 2021 made more than RM1 million in sales in their first year on the platform. Some of these sellers came from areas other than state capitals and major cities, such as Hutan Melintang in Perak, Kulim in Kedah, and Bera in Pahang, and they sold mobile and accessories, home and living, and health and beauty products, among other things.
2.4 Shoppe home made products
With a touch of irony, the tech-driven ecommerce industry is one of the best markets for old-fashioned, handcrafted crafts. More online shoppers are returning to handmade products and rustic styles, which are emerging as the iconic look of the late 2010s and early 2020s. From Shopee website, there are plenty of homemade products sold such as tarts, spices, gift box suprises, cookies, bath bombs, candles and many more variety types of things that are sold in one platform.
2.5 Analysis on Customer Feedback
Customer feedback is information, perspectives, concerns, and input provided by your community regarding their interactions with your company, product, or services. This feedback guides improvements to the customer experience and, also when negative, can empower positive change in any business. Customer feedback is critical for guiding and informing decisions, as well as influencing innovations and adjustments to your product and service. It is also necessary for determining customer satisfaction among current customers. Understanding how customers perceive your product, support, and company is priceless. Collecting and acting on customer feedback is essential for any company that wants to provide its customers with the products they require.(NeetuNarwal, 2018)did a study on using feedbacks to provide customer sentiment regarding product and concluded that It has been determined that social media has become an important and dependable source of public opinion in terms of people's trust and faith in any product.
2.6 Sentiment Analysis on Customer Feedback
Sentiment analysis is a Natural Language Processing (NLP) technique that involves extracting emotions from raw texts. This is typically used on posts on social media as well as customer reviews to determine whether some users are positively or negatively or even why. This section discusses previous literature as well as various sentiment analysis techniques. The overall tone of the reviews is starting with paper from (Sasikala & Mary Immaculate Sheela, 2020) An improved adaptive neuro fuzzy inference system methodology is proposed for future prediction of online product. The performance of both proposed methods is evaluated. The proposed Deep learning modified neural network is used for three review analysis scenarios: grade, content, and collaboration. The performance measures of ps, rk, fs, and ac are compared for data sets ranging from 1000 to 5000. Comparing the three scenarios, the collaboration-based scenario yields the best results for product analysis. Compared to the existing ANFIS, the proposed Improved adaptive neuro fuzzy inference system achieves the highest values for ps, rk, fs, and ac. Hence, the proposed Collaboration based scenario and improved adaptive neuro fuzzy inference system performed well for Sentiment Analysis and future prediction of online products. The system's flaw is that it only identifies the sentiment expressed in a word, but not the entire context of the piece. The proposed system can be improved in the future by solving the keyword processing problem and using a hybridization algorithm in the future prediction process. Followed byThis paper (Tsvetanka L. Georgieva-Trifonova et al., n.d.) investigates text classification and its use for online store customer feedback. A dataset of user reviews in Bulgarian is created and published for this purpose. A model that extracts pointwise mutual information of words in relation to categories is also proposed. The results of the experiments confirm the proposed model's utility in various classifier applications such as SVM (Support Vector Machine), SVM (Support Vector Machine), Naive Bayes, H2O’s Deep Learning, Rule-based classifiers Ridor, Jrip, PART. In future work, text analysis of customer reviews will include emoticons and words written entirely in capital letters. The next (Zhao et al., 2021)interestingly proposed a model. This paper provides quantified satisfaction functions based on online reviews from the perspective of risk attitude and aspiration. Here are the main innovations. First, this paper proposes an improved Kano model based on online comments to identify attribute types, which are then illustrated from a risk attitude perspective. Previous online review-based Kano model studies identified attribute types but did not analyse their impact on consumer satisfaction from a risk attitude perspective. Second, this paper mines consumer aspirations based on attribute risk attitudes and their impact on consumer satisfaction. The quantified satisfaction functions are constructed to provide more objective and accurate improvement suggestions. They only analyse the relationship among attribute performance and customer satisfaction qualitatively and make vague improvement suggestions. Third, three weight curves that can reflect consumer emphasis on positive and negative feedback under different attribute types are proposed. As a result, various weight curves are used to aggregate the single feedback for different feature types. Finally, this paper not only distinguishes positive and negative sentiments in online reviews, but also quantifies them. Existing Kano model studies on online feedback only identified positive and negative sentiments, but not their degrees. This paper also has a flaw. However, this paper only considers positive and negative sentiments of attributes, not the frequency with which consumers refer to them. (Jain et al., 2021) A cuckoo optimised machine learning model is proposed to predict airline recommendations. The data was scraped from https://www.airlinequality.com. The proposed CS-XGB classifier outperforms other state-of-the-art techniques.It will be interesting to see how different methods perform for different service aspects in the future. An airline's logo, images, or origin may be able to explain consumer sentiments more effectively than the current study. It is possible to demonstrate consumer attitude in future research. Consumer sentiments in the airline industry were studied. Similarly, our proposed model can be used in various real-life applications such as hotels, restaurants, tourism destinations, online services, etc. In future research, researchers may combine cultural variables with online reviews to explain and predict consumer recommendations. We suggest modifying CS approaches and implementing them on deep learning models like Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) to improve recommendation prediction accuracy and robustness. This review(Ain et al., 2017) highlights recent research on using deep learning models like deep neural networks and convolutional neural networks to solve problems like sentiment classification, cross-lingual problems, textual and visual analysis, and product review analysis. Using deep learning methods for sentiment analysis can be more efficient and accurate, according to these studies. Since deep learning models predict or mimic the human mind, they are more accurate than shallow models. Deep learning networks outperform SVMs and normal neural networks because they have more hidden layers than SVMs and normal neural networks. Deep learning networks can train both supervised and unsupervised. Deep learning networks automatically extract features without human intervention, saving time by eliminating the need for feature engineering. Sentiment Analysis has various problem statements. The ability to adapt to task changes with minimal system changes is a strength of Deep Learning. Unlike earlier models like SVM, this method has some limitations. It requires a lot of data and is expensive to train. These complex models can take weeks to train on expensive GPUs. Last but not least, in this paper (Yi & Liu, 2020) a hybrid recommendation system that would self-study customer shopping data in the form of reviews, understand the pattern and predict customer interest in buying a particular product in a selected shop was developed. The Hybrid Recommendation System (HRS) was intended using a machine learning based regression model. This framework has been proven to be helpful in categorising customers' preferred shops based on their purchases. This HRS approach has the advantage of requiring no human intervention in predicting customer shopping preferences. The efficiency of this Hybrid Recommendation System was evaluated using three metrics: Mean absolute error, mean squared error, and mean absolute percentage error. The results discussed previously show that HRS has a lower MAE than the other approaches compared. Also, the MAPE value for HRS was nearly 98 percent, indicating high accuracy. Similarly, the MSE value for HRS was low, indicating high precision and accuracy. On the basis of accurate prediction of customer sentiment towards a particular shop, the proposed Recommendation System (HRS) help performance other contemporary approaches. This approach can be extended to gather customer interest in multiple products across multiple locations.
In this section will explain the overall process of the research that will be done, in the first step, will start with data collection from Shopee Malaysia customer feedback specifically only on home-based product. Where data is extracted from Shopee Malaysia. The second phase data will be cleaned before moving on to pre-processing. Next will be the data pre-processing phase and prepare the data to be used for model development. The third phase is applying lexicon on processed data lastly, we will analyse the result till to an acceptable degree of accuracy.
3.2 RESEARCH SCOPE
The scope of the research is limited to only home-based products and focusing only on one country and only one online platform which is Shopee Malaysia. Data will be extracted specially from sellers who sells home-based products only.
 Global Data (2020, September 8) COVID-19 accelerates e-commerce growth in Malaysia, says GlobalData. https://www.globaldata.com/covid-19-accelerates-e-commerce-growth-malaysia-says-globaldata/
 incfile (2021, April 8), Home-Based Entrpreneur Statistics to Know in 2021 https://www.incfile.com/blog/post/shocking-us-home-based-business-statistics
 International Trade Administration (2021, June 11), Describes what a company needs to know to take advantage of e-commerce in the local market and covers prominent B2B websites. https://www.trade.gov/country-commercial-guides/malaysia-ecommerce
To get code implementation with research paper report writing project then you can share your project requirement details at: