Success Prediction on Crowdfunding with Multimodal Deep Learning
Authors: Chaoran Cheng, Fei Tan, Xiurui Hou, Zhi Wei
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our extensive experimental results show that the image features could improve success prediction performance significantly, particularly for project profiles with little text information. |
| Researcher Affiliation | Academia | Department of Computer Science, New Jersey Institute of Technology, USA {cc424, ft54, xh256, zhi.wei}@njit.edu |
| Pseudocode | No | The paper describes the model architecture and processes in prose and mathematical formulations, but it does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The data and code will be released at https://github.com/andrea1980s/Kickstarter |
| Open Datasets | Yes | The evaluation of our MDL framework is done on the dataset scraped from Kickstarter. We crawled the data using the seed from webrobots5. We ran the scraping script for 2 weeks to collect the data. The data and code will be released at https://github.com/andrea1980s/Kickstarter |
| Dataset Splits | Yes | To train our model, select best parameters and evaluate the performance, we split the dataset into 3 parts as shown in Table 1... To evaluate the predictive potential on shifting distribution, we set campaigns launched at 2015 and 2016 as the training set, campaigns at 2017 as the validation set, and campaigns at 2018 as the testing set. |
| Hardware Specification | No | The paper describes software frameworks and training parameters (e.g., 'implemented in Python using Keras with Tensor Flow backend'), but it does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper states, 'Our MDL model is implemented in Python using Keras with Tensor Flow backend.' However, it does not specify version numbers for Python, Keras, or TensorFlow, which are necessary for reproducible software dependencies. |
| Experiment Setup | Yes | The models are tuned based on validation part and the optimal parameters are reported accordingly based on F1 measure. Our MDL model is implemented in Python using Keras with Tensor Flow backend. We used the RMSprop [Tieleman and Hinton, 2012] optimizer and the learning rate is set to 1e5 for 100 epochs. We set the batch size to 128 campaign projects and employed early stopping with 20 epochs and dropout [Srivastava et al., 2014] to prevent overfitting. |