ts-imputation

Time SeriesTime-Series-Library

Description

Time Series Imputation: Custom Model Design

Objective

Design and implement a custom deep learning model for time series missing value imputation. Your code goes in the Model class in models/Custom.py. Three reference implementations (DLinear, TimesNet, PatchTST) are provided as read-only.

Evaluation

Trained and evaluated on three multivariate datasets with 25% random masking:

ETTh1 (7 variables)
Weather (21 variables)
ECL (321 variables)

All use seq_len=96. Metrics: MSE and MAE on masked regions only (lower is better).

Code

Custom.py

EditableRead-only

1import torch
2import torch.nn as nn
3
4
5class Model(nn.Module):
6    """
7    Custom model for time series imputation.
8
9    Forward signature: forward(x_enc, x_mark_enc, x_dec, x_mark_dec, mask=None)
10    - x_enc: [batch, seq_len, enc_in] — input with masked values set to 0
11    - x_mark_enc: [batch, seq_len, time_features] — time feature encoding
12    - x_dec: not used for imputation (None)
13    - x_mark_dec: not used for imputation (None)
14    - mask: [batch, seq_len, enc_in] — binary mask (1=observed, 0=masked)
15

Additional context files (read-only):

Time-Series-Library/models/DLinear.py
Time-Series-Library/models/TimesNet.py
Time-Series-Library/models/PatchTST.py
Time-Series-Library/layers/AutoCorrelation.py
Time-Series-Library/layers/Autoformer_EncDec.py
Time-Series-Library/layers/Conv_Blocks.py
Time-Series-Library/layers/Crossformer_EncDec.py
Time-Series-Library/layers/Embed.py
Time-Series-Library/layers/FourierCorrelation.py
Time-Series-Library/layers/SelfAttention_Family.py
Time-Series-Library/layers/StandardNorm.py
Time-Series-Library/layers/Transformer_EncDec.py

Results

Model	Type	mse ETTh1 ↓	mae ETTh1 ↓	mse Weather ↓	mae Weather ↓	mse ECL ↓	mae ECL ↓
dlinear	baseline	0.150	0.269	0.050	0.114	0.113	0.243
patchtst	baseline	0.104	0.216	0.031	0.052	0.062	0.169
patchtst	baseline	0.106	0.217	0.031	0.053	0.062	0.169
timesnet	baseline	0.080	0.191	0.029	0.055	0.091	0.208
anthropic/claude-opus-4.6	vanilla	0.037	0.131	0.051	0.092	0.078	0.192
deepseek-reasoner	vanilla	0.039	0.130	0.029	0.055	0.076	0.190
google/gemini-3.1-pro-preview	vanilla	0.064	0.171	0.109	0.169	0.795	0.743
openai/gpt-5.4-pro	vanilla	0.042	0.138	0.026	0.038	0.055	0.157
qwen/qwen3.6-plus:free	vanilla	0.188	0.281	0.054	0.080	0.118	0.239
anthropic/claude-opus-4.6	agent	0.037	0.129	0.031	0.058	0.077	0.192
deepseek-reasoner	agent	-	-	-	-	-	-
deepseek-reasoner	agent	0.039	0.130	0.029	0.055	0.076	0.190
google/gemini-3.1-pro-preview	agent	0.084	0.193	0.043	0.068	0.050	0.148
openai/gpt-5.4-pro	agent	0.044	0.141	0.025	0.036	0.050	0.150
qwen/qwen3.6-plus:free	agent	0.188	0.281	0.054	0.080	0.118	0.239

Agent Conversations

anthropic/claude-opus-4.6

6 steps

deepseek-reasoner

6 steps

google/gemini-3.1-pro-preview

7 steps

openai/gpt-5.4-pro

6 steps