UTS Metodologi dan Penulisan Ilmiah 2025

UTS Metodologi Penelitian dan Penulisan Ilmiah 2024/2025

Senin, 14 April 2025 · 13.00–14.40 WIB (100 menit) · Buku tertutup. Jawab pada lembar yang disediakan.

1 (20%)

Jelaskan bagaimana sebaiknya latar belakang proposal penelitian disusun.

2 (20%)

Sebutkan dan jelaskan minimal empat karakteristik penelitian.

3 (20%)

Sebutkan dua jenis eligible data source dan dua jenis not eligible data source pada literature review. Jelaskan alasan masing-masing.

4 (20%)

Perhatikan potongan artikel berikut. Buat kalimat literature review untuk dua dari empat kategori compare, contrast, critisize, synthesize. Tuliskan kategori dan sitasi APA versi terakhir, dalam bahasa Indonesia. Bebas memilih judul dan jumlah artikel referensi per kategori.

Exploring the Effectiveness of Agile and Traditional Requirement Elicitation Techniques

Martinez, A., & Lee, S. (2023). International Journal of Software Engineering and Knowledge Engineering, 33(5), 789–805.

Agile and traditional (waterfall) software development technology aims to accurately capture user requirements to ensure the developed software meets stakeholder needs. Both methodologies involve stakeholders in the requirement-gathering process to gather necessary information. The Agile methods employ iterative cycles, allowing continuous refinement of requirements; thus, these methodologies have higher process flexibility than traditional methodologies that follow a linear approach with the defined upfront requirements. In addition, the traditional methods emphasize comprehensive documentation during the initial phases, while Agile prioritizes working software and minimal documentation.

Agile vs. Traditional Requirement Elicitation Methods: A Comparative Study

Chen, L., & Kumar, R. (2022). Proceedings of the International Conference on Software Engineering and Applications, 150–160.

Agile and traditional (waterfall) methods aim to accurately capture and document software requirements to ensure the final product meets stakeholder needs. Both approaches involve stakeholders in the requirement elicitation process to varying degrees to gather necessary information. Traditional methods follow a linear and sequential process, emphasizing comprehensive documentation upfront, whereas agile methods utilize iterative cycles with continuous stakeholder feedback. Agile approaches encourage ongoing and frequent stakeholder interactions, fostering a collaborative environment. Traditional methods typically involve stakeholder input, primarily during the initial requirement-gathering phase.

Sentiment Analysis of Persian Movie Reviews Using Deep Learning

Kia Dashtipour, Mandar Gogate, Ahsan Adeel, Hadi Larijani, Amir Hussain. Entropi, Vol.23, Issue5, May2021, pp.1–16.

Sentiment analysis aims to automatically classify the subject’s sentiment (e.g., positive, negative, or neutral) towards a particular aspect such as a topic, product, movie, news, etc. Deep learning has recently emerged as a powerful machine learning technique to tackle the growing demand for accurate sentiment analysis. However, the majority of research efforts are devoted to English-language only, while information of great importance is also available in other languages. This paper presents a novel, context-aware, deep-learning-driven, Persian sentiment analysis approach. Specifically, the proposed deep-learning-driven automated feature-engineering approach classifies Persian movie reviews as having positive or negative sentiments. Two deep learning algorithms, convolutional neural networks (CNN) and long-short-term memory (LSTM), are applied and compared with our previously proposed manual-feature-engineering-driven, SVM-based approach. The simulation results demonstrate that deep learning outperformed state-of-the-art, shallow machine learning approaches. Specifically, the stacked-bidirectional-LSTM achieved the highest accuracy, of up to 95.61%, for the movie dataset, but not for the hotel dataset, where 2D-CNN did better (89.76%). Furthermore, we have evaluated the best performance of deep learning classifiers with different layers. Ongoing and future work aims to extend the developed corpus to multilingual product reviews.

Word2Vec for Indonesian Sentiment Analysis towards Hotel Reviews: An Evaluation Study

Rizka Putri Nawangsari, Retno Kusumaningrum, Adi Wibowo. The 4th International Conference on Computer Science and Computational Intelligence, 2019, pp.728–735.

There are several conducted studies of sentiment analysis tasks for the Indonesian language, such as sentiment analysis towards movie reviews, tweets of social media twitter, sales reviews of the marketplace, hotel reviews, etc. Most of those studies implement the baseline shallow learning methods such as MaxEnt (Maximum Entropy), Naive Bayes, and Support Vector Machine (SVM). Furthermore, those studies generally employ hand-crafted features such as word occurrence, word presence, TF-IDF, sentiment lexicon, etc. The performance of many studies which implement combination between shallow learning methods and hand-crafted features commonly depends on the selected of data representation or features. In other words, the performance of shallow learning methods depends on the success of hand-crafted features. It is caused by an inability of shallow learning methods to extract and organize discriminative information from data. Therefore, the implementation of shallow learning methods always requires feature engineering process. Unfortunately, feature engineering process has several drawbacks, i.e. (i) It is labor-intensive, (ii) It highly-cost to obtain the preferred accuracy since it needs manual pre-processing, e.g. POS tagging and stemming, that is time-consuming and challenging task.

Sentiment Analysis using Word2Vec and Long Short-Term Memory (LSTM) for Indonesian Hotel Reviews

Putra Fissabil Muhammad, Retno Kusumaningrum, Adi Wibowo. The 5th International Conference on Computer Science and Computational Intelligence, 2020, pp.728–735.

The classical machine learning methods have a problem wherein it is difficult to determine in the feature extraction what to be included in the given model. If features are missing or incomplete, the model will generate imperfect outcomes. If there are too many features and if all of them do not contribute to the output of the model, this model will not give optimal performance. Nonetheless, it has been found that this problem can be overcome by using CNN (Convolutional Neural Network) since it is a deep learning technique that uses the neural network framework where every layer obtains input from the preceding layer and is delivered to the subsequent layer. Nonetheless, CNN also has the drawback that it cannot work with long sequential data. It is because CNN does not possess a memory, so it cannot retain information about the word meaning. This drawback can be overcome by using the LSTM (Long-Short Term Memory) model. LSTM is a kind of RNN (Recurrent Neural Network) architecture which is designed to “retain” values that have been obtained before for a specific period. LSTM consists of 3 gates that regulate the flow, which are the input gate, forget gate, as well as output gate. The input gate regulates the input of new data into memory, the forget gate regulates how long particular values get stored in memory, and the output gate regulates how much the value retained in memory influences the output block activation [9].

5 (20%)

Jelaskan experimental research, prosedurnya, dan contohnya.