2025年度ベンチャー体験工房3
Undergraduate School and Graduate School
学部?大学院
- ベンチャー体験工房3
- Developing a Novel Pattern Mining Model to Discover Hidden Patterns in Fukushima Traffic Congestion Big Data
- - Japan Road Transportation Information Center (JARTIC) has set up the sensor network to monitor traffic congestion in Fukushima.
- Each road-segment in this network generates data at every 5-minute interval.
- Previous year, we have developed a data warehouse technology that generates data frames at 10 times faster than the state-of-the-art.
- This year, we plan to develop a novel pattern algorithm to discover hidden patterns. - シラバス詳細
2025年4月
4月7日
Introduction to Traffic Information Systems: Understanding JARTIC and TCSS
授業内容：We covered the role of the Japan Road Traffic Information Center (JARTIC) and the Traffic Congestion Statistical System (TCSS). We explored how JARTIC collects and distributes traffic information through 133 centers nationwide. The importance of real-time traffic data for road users was highlighted.
4月14日
Data Collection Techniques: Sensors and Measurement Points
授業内容：We learned about the data collection process using sensors installed at over 40,000 measurement points. The types of sensors, such as ultrasonic vehicle detectors, and the significance of collecting traffic volume and occupancy time every 5 minutes were thoroughly analyzed.
4月21日
Real-Time Traffic Analysis: Interpreting Congestion Data
授業内容：This class focused on the processing and interpretation of live traffic data, including the classification of congestion status into traffic jam, congestion, no congestion, and unknown (sensor abnormality). We practiced accessing this data through the TCSS interface.
2025年5月
5月5日
授業内容： Temporal and Spatial Resolution of TCSS Data :
Analyzes how TCSS offers high-resolution temporal (5-minute intervals) and spatial (road link-level and mesh code) data. This granularity allows researchers to study traffic dynamics over time and space with precision.
5月12日
授業内容：Congestion Detection and Classification Algorithms in TCSS:
Focuses on the definitions and criteria used to detect and classify congestion. It explains how TCSS uses speed thresholds and vehicle detection data to classify congestion into levels such as "light" or "heavy" across various road types.
5月19日
授業内容：Visualization Techniques: Mapping and Graphical Representation in TCSS.
Details the system's ability to visualize traffic data through interactive maps, Excel tables, and statistical graphs. It discusses how these visualization features support both operational monitoring and academic research.
2025年6月
6月2日
授業内容：Traffic Data Normalization and Anomaly Handling:
We removed unnecessary unnamed columns, detected abnormal traffic values, and applied threshold-based normalization to standardize the dataset for modeling.
6月9日
授業内容：Missing Value Imputation in Traffic Datasets:
We explored multiple imputation techniques--mean, median, mode, KNN, forward fill, and backward fill--to handle missing data. After evaluation, mean imputation was selected as the best method.
6月16日
授業内容：Deep Learning Models for Traffic Forecasting:
We experimented with various time-series deep learning models including LSTM, Bi-LSTM, GRU, Autoencoders, Transformers, and CNNs to predict traffic patterns.
6月23日
授業内容：LSTM-Based Traffic Flow Prediction:
We implemented an LSTM model that performed well due to its layered architecture and ability to capture long-term dependencies in traffic data.
