In deep learning, the technique used to skip one or more layers by connecting non-adjacent layers is called "Skip Connections." Skip connections allow the model to bypass one or more layers and facilitate the flow of information from one layer to another, helping in the training of deep neural networks.

Time-series data often has specific characteristics, such as seasonality and trends. LightGBM is well-suited for such data as it can handle categorical features efficiently and is capable of capturing complex patterns, making it a strong choice for time-series forecasting.

An event-driven database, such as Apache Kafka, is suitable for capturing and storing real-time activities and interactions, especially when the data is unstructured, and the schema might evolve over time.

The primary goal of EDA is to summarize and explore data. It involves visualizing and understanding the dataset's main characteristics and relationships before diving into more advanced tasks, such as model building or predictions. EDA helps identify patterns and anomalies in the data.

The primary characteristic that differentiates Big Data from traditional datasets is "Variety." Big Data often includes a wide range of data types, including structured, unstructured, and semi-structured data, making it more diverse than traditional datasets.

"Data-driven decision-making" underscores the significance of using data to inform decisions. It implies that decisions should be backed by data and analysis rather than relying solely on intuition. This approach enhances the quality and reliability of decision-making.

Recall is particularly useful when dealing with imbalanced datasets because it measures the ability of a model to identify all relevant instances of a class. In such scenarios, accuracy can be misleading, as the model may predict the majority class more frequently, resulting in a high accuracy but poor performance on the minority class. Recall, also known as true positive rate, focuses on capturing as many true positives as possible.

ARIMA is a time series forecasting method that utilizes past observations to predict future values. It incorporates autoregressive and moving average components, making it suitable for analyzing time series data. The other options are not specifically designed for time series forecasting and do not rely on past observations in the same way.

Unsupervised Learning is the type where the model discovers patterns or structures without prior data labeling. Common tasks in this category include clustering and dimensionality reduction, helping find hidden insights in data without any guidance.

Regularization techniques add a "Weight penalty" term to the loss function to constrain the magnitude of the model parameters, preventing them from becoming excessively large. This helps prevent overfitting and improves the model's generalization capabilities. Regularization is a crucial concept in machine learning and deep learning.