Abstract:Missing data is a pervasive challenge in wireless networks and many other domains, often compromising the performance of machine learning and deep learning models. To address this, we propose a novel framework, FGATT, that combines the Fuzzy Graph Attention Network (FGAT) with the Transformer encoder to perform robust and accurate data imputation. FGAT leverages fuzzy rough sets and graph attention mechanisms to capture spatial dependencies dynamically, even in scenarios where predefined spatial information is unavailable. The Transformer encoder is employed to model temporal dependencies, utilizing its self-attention mechanism to focus on significant time-series patterns. A self-adaptive graph construction method is introduced to enable dynamic connectivity learning, ensuring the framework's applicability to a wide range of wireless datasets. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods in imputation accuracy and robustness, particularly in scenarios with substantial missing data. The proposed model is well-suited for applications in wireless sensor networks and IoT environments, where data integrity is critical.
What problem does this paper attempt to address?
This paper attempts to solve the ubiquitous problem of data missing in wireless networks. Specifically, data missing can seriously affect the performance of machine - learning and deep - learning models, especially in wireless sensor networks and the Internet of Things (IoT) environment where data integrity is crucial. To solve this problem, the author proposes a new framework - FGATT (Fuzzy Graph Attention Networks and Transformer Encoders), which combines the Fuzzy Graph Attention Network (FGAT) and Transformer encoders to achieve robust and accurate data imputation.
### Main problems:
1. **The impact of data missing on model performance**: Data missing can lead to a decline in the performance of machine - learning and deep - learning models because most algorithms require complete data for training and inference.
2. **Limitations of existing methods**:
- **Statistical methods**: Methods such as ARIMA, SVR, and LR are computationally efficient and easy to interpret, but they perform poorly when dealing with non - linear and large - scale datasets.
- **Time - series deep - learning methods**: Models such as FFN, LSTM, GRU, and Transformer can well capture time - series patterns, but often ignore spatial dependencies, which are crucial in many practical applications.
3. **Lack of spatial information**: In wireless sensor networks, the spatial relationships between sensors have a great impact on the quality and accuracy of data imputation, and existing methods are usually unable to effectively utilize this spatial information.
### Solutions:
- **Dynamic graph construction**: Introduce an adaptive connectivity learning method, which does not require a predefined spatial structure and can dynamically learn the connection relationships between nodes.
- **Hybrid framework design**: Combine FGAT for capturing spatial dependencies and Transformer encoders for modeling time dependencies, thereby achieving comprehensive modeling of spatio - temporal dependencies.
- **Application of fuzzy rough sets**: By integrating fuzzy rough sets with the graph attention mechanism, FGAT can capture more powerful feature representations in uncertain spatial relationships.
### Experimental results:
Through extensive experimental verification, FGATT is significantly superior to the existing state - of - the - art methods in terms of imputation accuracy and robustness, especially in the case of a large amount of data missing. This model is suitable for data imputation tasks in wireless sensor networks and IoT environments, ensuring data integrity and reliability.
### Formula display:
1. **Fuzzy lower approximation and upper approximation**:
\[
R_B^d(x)=\inf_{y\in U}\max(1 - R(x,y),d(y))
\]
\[
R_B^d(x)=\sup_{y\in U}\min(R(x,y),d(y))
\]
2. **Connectivity score**:
\[
\text{Score}_t(i,j)=\alpha\times R_B^d_j(x_t^i)+(1 - \alpha)\times R_B^d_i(x_t^j)
\]
3. **Aggregated connectivity score**:
\[
\text{Score}(i,j)=\frac{1}{T}\sum_{t = 1}^{T}\text{Score}_t(i,j)
\]
4. **Graph attention mechanism**:
\[
h'_i=\text{LeakyReLU}\left(\sum_{j\in N(i)}\alpha_{ij}W h_j\right)
\]
5. **Layer normalization**:
\[
y=\frac{x-\mu}{\sqrt{\sigma^2+\epsilon}}\cdot\gamma+\beta
\]
6. **Self - attention mechanism**:
\[
\text{Attention}(Q,K,V)=\text{soft}