This dataset provides a comparison of various Large Language Models (LLMs) based on their performance, cost, and efficiency. It includes important details like speed, latency, benchmarks, and pricing, helping users understand how different models stack up against each other.

Key Details:

• File Name: llm_comparison_dataset.csv
• Size: 14.57 kB
• Total Columns: 15
• License: CC0 (Public Domain)

What’s Inside?

Here are some of the key metrics included in the dataset:

1. Context Window: Maximum number of tokens the model can process at once.
2. Speed (tokens/sec): How fast the model generates responses.
3. Latency (sec): Time delay before the model responds.
4. Benchmark Scores: Performance ratings from MMLU (academic tasks) and Chatbot Arena (real-world chatbot performance).
5. Open-Source: Indicates if the model is publicly available or proprietary.
6. Price per Million Tokens: The cost of using the model for one million tokens.
7. Training Dataset Size: Amount of data used to train the model.
8. Compute Power: Resources needed to run the model.
9. Energy Efficiency: How much power the model consumes.

This dataset is useful for researchers, developers, and AI enthusiasts who want to compare LLMs and choose the best one based on their needs.

📌If you find this dataset useful, do give an upvote :)

📊 Student Performance Dataset (Synthetic, Realistic)

Overview

This dataset contains 1000000 rows of realistic student performance data, designed for beginners in Machine Learning to practice Linear Regression, model training, and evaluation techniques.

Each row represents one student with features like study hours, attendance, class participation, and final score.
The dataset is small, clean, and structured to be beginner-friendly.

🔑 Columns Description

• student_id → Unique identifier for each student.
• weekly_self_study_hours → Average weekly self-study hours (0–40). Generated using a normal distribution centered around 15 hours.
• attendance_percentage → Attendance percentage (50–100). Simulated with a normal distribution around 85%.
• class_participation → Score between 0–10 indicating how actively the student participates in class. Generated from a normal distribution centered around 6.
• total_score → Final performance score (0–100). Calculated as a function of study hours + random noise, then clipped between 0–100. Stronger correlation with study hours.
• grade → Categorical label (A, B, C, D, F) derived from total_score.

📐 Data Generation Logic

1. Weekly Study Hours: Modeled using a normal distribution (mean ≈ 15, std ≈ 7), capped between 0 and 40 hours.
2. Scores: More study hours → higher score. Formula:

Random noise simulates differences in learning ability, motivation, etc.

1. Attendance & Participation: Independent but realistic variations added.
2. Grades: Assigned from scores using thresholds:

• A: ≥ 85
• B: ≥ 70
• C: ≥ 55
• D: ≥ 40
• F: < 40

🎯 How to Use This Dataset

Regression Tasks

• Predict total_score from weekly_self_study_hours.
• Train and evaluate Linear Regression models.
• Extend to multiple regression using attendance_percentage and class_participation.

Classification Tasks

• Predict grade (A–F) using study hours, attendance, and participation.

Model Evaluation Practice

• Apply train-test split and cross-validation.
• Evaluate with MAE, RMSE, R².
• Compare simple vs. multiple regression.

✅ This dataset is intentionally kept simple, so that new ML learners can clearly see the relationship between input features (study, attendance, participation) and output (score/grade).

In situations where the cost/benefit analysis of using physics-based damage propagation algorithms is not favorable and when sufficient test data are available that map out the damage space, one can employ data-driven approaches. In this investigation, we evaluate different algorithms for their suitability in those circumstances. We are interested in assessing the trade-off that arises from the ability to support uncertainty management, and the accuracy of the predictions. We compare here a Relevance Vector Machine (RVM), Gaussian Process Regression (GPR), and a Neural Network-based approach and employ them on relatively sparse training sets with very high noise content. Results show that while all methods can provide remaining life estimates although different damage estimates of the data (diagnostic output) changes the outcome considerably. In addition, we found that there is a need for performance metrics that provide a comprehensive and objective assessment of prognostics algorithm performance.

The current dataset is consisted of 200 search results extracted from Google and Bing engines (100 of Google and 100 of Bing). The search terms are selected from the 10 most search keywords of 2021 based on the provided data of Google Trends. The rest of the sheets include the performance of the websites according to three technical evaluation aspects. That is, SEO, Speed and Security. The performance dataset has been developed through the utilization of CheckBot crawling tool. The whole dataset can help information retrieval scientists to compare the two engines in terms of their position/ranking and their performance related to these factors.

For more information about the thinking of the of the structure of the dataset please contact the Information Management Lab of University of West Attica.

Contact Persons: Vasilis Ntararas (lb17032@uniwa.gr) , Georgios Ntimo (lb17100@uniwa.gr) and Ioannis C. Drivas (idrivas@uniwa.gr)

The dataset used in the paper is a new dataset for zero-shot performance comparison of spatial conditions.

The top table shows the average classifier performance for cross-validation on the 9-locus public STR data. The bottom table is the performance for the same test, but on a 9-locus subset of our ground-truth training data. While overall performance is lower than the 15-locus cross-validation test on our ground-truth data (Table 1), the two data sets perform similarly here, indicating that increasing the number of markers in the data set can significantly improve performance.

Context

This dataset contains campaign data for both Facebook Ads and AdWords, offering a side-by-side comparison of performance metrics, costs, and conversions. It's an ideal resource for A/B testing in marketing, especially for analyzing the effectiveness of ad campaigns across platforms.

Source

This dataset was created from scratch using Mockaroo, ensuring it is tailored for practical use.

Inspiration

While watching a YouTube tutorial 👉 [https://youtu.be/iCj4lT5KvJk?si=FijILsrbxBrcE3pw])(url), I noticed that the tutorial lacked an uploaded dataset, and many viewers in the comment section requested one. To help others follow along and practice, I decided to create a mock dataset from scratch. Now, you can easily replicate the tutorial and enhance your skills!

Analysis Ideas

Platform Performance Comparison: Compare key metrics like CTR, conversion rate, and cost per click between Facebook Ads and AdWords.

Trend Over Time: Analyze changes in ad performance metrics across different years.

A/B Testing Insights: Assess simultaneous campaigns to identify the better-performing platform.

Cost Efficiency: Identify campaigns with low costs but high conversions on each platform.

Visualization of Metrics: Create charts to visually compare campaign performance. Statistical Insights: Perform hypothesis testing to check for significant differences in performance metrics. Recommendations for Marketing Strategy: Provide actionable suggestions based on the data analysis. # Enjoy exploring and testing this dataset for your marketing analyses!

Performance comparison of each model on the test dataset.

The receiver operating characteristics (ROC) curve is typically employed when one wants to evaluate the discriminatory capability of a continuous or ordinal biomarker in the case where two groups are to be distinguished, commonly the ’healthy’ and the ’diseased’. There are cases for which the disease status has three categories. Such cases employ the (ROC) surface, which is a natural generalization of the ROC curve for three classes. In this paper, we explore new methodologies for comparing two continuous biomarkers that refer to a trichotomous disease status, when both markers are applied to the same patients. Comparisons based on the volume under the surface have been proposed, but that measure is often not clinically relevant. Here, we focus on comparing two correlated ROC surfaces at given pairs of true classification rates, which are more relevant to patients and physicians. We propose delta-based parametric techniques, power transformations to normality, and bootstrap-based smooth nonparametric techniques to investigate the performance of an appropriate test. We evaluate our approaches through an extensive simulation study and apply them to a real data set from prostate cancer screening.

Model performance comparison on the VOC2012 dataset.

The Google Reviews & Ratings Dataset provides businesses with structured insights into customer sentiment, satisfaction, and trends based on reviews from Google. Unlike broad review datasets, this product is location-specific—businesses provide the locations they want to track, and we retrieve as much historical data as possible, with daily updates moving forward.

This dataset enables businesses to monitor brand reputation, analyze consumer feedback, and enhance decision-making with real-world insights. For deeper analysis, optional AI-driven sentiment analysis and review summaries are available on a weekly, monthly, or yearly basis.

Dataset Highlights

• Location-Specific Reviews – Reviews and ratings for the locations you provide.
• Daily Updates – New reviews and rating changes updated automatically.
• Historical Data Access – Retrieve past reviews where available.
• AI Sentiment Analysis (Optional) – Summarized insights by week, month, or year.
• Competitive Benchmarking – Compare performance across selected locations.

Use Cases

• Franchise & Retail Chains – Monitor brand reputation and performance across locations.
• Hospitality & Restaurants – Track guest sentiment and service trends.
• Healthcare & Medical Facilities – Understand patient feedback for specific locations.
• Real Estate & Property Management – Analyze tenant and customer experiences through reviews.
• Market Research & Consumer Insights – Identify trends and analyze feedback patterns across industries.

Data Updates & Delivery

• Update Frequency: Daily
• Data Format: CSV for easy integration
• Delivery: Secure file transfer (SFTP or cloud storage)

Data Fields Include:

• Business Name
• Location Details
• Star Ratings
• Review Text
• Timestamps
• Reviewer Metadata

Optional Add-Ons:

• AI Sentiment Analysis – Aggregate trends by week, month, or year.
• Custom Location Tracking – Tailor the dataset to fit your specific business needs.

Ideal for

• Marketing Teams – Leverage real-world consumer feedback to optimize brand strategy.
• Business Analysts – Use structured review data to track customer sentiment over time.
• Operations & Customer Experience Teams – Identify service issues and opportunities for improvement.
• Competitive Intelligence – Compare locations and benchmark against industry competitors.

Why Choose This Dataset?

• Accurate & Up-to-Date – Daily updates ensure fresh, reliable data.
• Scalable & Customizable – Track only the locations that matter to you.
• Actionable Insights – AI-driven summaries for quick decision-making.
• Easy Integration – Delivered in a structured format for seamless analysis.

By leveraging Google Reviews & Ratings Data, businesses can gain valuable insights into customer sentiment, enhance reputation management, and stay ahead of the competition.

Dataset for journal article "Comparing the performance and coverage of selected in silico (liver) metabolism tools relative to reported studies in the literature to inform analogue selection in read-across: A case study". This dataset is associated with the following publication: Boyce, M., B. Meyer, C. Grulke, L. Lizarraga, and G. Patlewicz. Comparing the performance and coverage of selected in silico (liver) metabolism tools relative to reported studies in the literature to inform analogue selection in read-across: A case study. Computational Toxicology. Elsevier B.V., Amsterdam, NETHERLANDS, 21: 100208, (2022).

This dataset contains GPS tracking data and performance metrics for motorcycle taxis (boda bodas) in Nairobi, Kenya, comparing traditional internal combustion engine (ICE) motorcycles with electric motorcycles. The study was conducted in two phases:Baseline Phase: 118 ICE motorcycles tracked over 14 days (2023-11-13 to 2023-11-26)Transition Phase: 108 ICE motorcycles (control) and 9 electric motorcycles (treatment) tracked over 12 days (2023-12-10 to 2023-12-21)The dataset is organised into two main categories:Trip Data: Individual trip-level records containing timing, distance, duration, location, and speed metricsDaily Data: Daily aggregated summaries containing usage metrics, economic data, and energy consumptionThis dataset enables comparative analysis of electric vs. ICE motorcycle performance, economic modelling of transportation costs, environmental impact assessment, urban mobility pattern analysis, and energy efficiency studies in emerging markets.Institutions:EED AdvisoryClean Air TaskforceStellenbosch UniversitySteps to reproduce:Raw Data CollectionGPS tracking devices installed on motorcycles, collecting location data at 10-second intervalsRider-reported information on revenue, maintenance costs, and fuel/electricity usageProcessing StepsGPS data cleaning: Filtered invalid coordinates, removed duplicates, interpolated missing pointsTrip identification: Defined by >1 minute stationary periods or ignition cyclesTrip metrics calculation: Distance, duration, idle time, average/max speedsDaily data aggregation: Summed by user_id and date with self-reported economic dataValidation: Cross-checked with rider logs and known routesAnonymisation: Removed start and end coordinates for first and last trips of each day to protect rider privacy and home locationsTechnical InformationGeographic coverage: Nairobi, KenyaTime period: November-December 2023Time zone: UTC+3 (East Africa Time)Currency: Kenyan Shillings (KES)Data format: CSV filesSoftware used: Python 3.8 (pandas, numpy, geopy)Notes: Some location data points are intentionally missing to protect rider privacy. Self-reported economic and energy consumption data has some missing values where riders did not report.CategoriesMotorcycle, Transportation in Africa, Electric Vehicles

A performance comparison between data structures.

The Building Performance Database (BPD) is the largest publicly-available source of measured energy performance data for buildings in the United States. It contains information about the building's energy use, location, and physical and operational characteristics. The BPD can be used by building owners, operators, architects and engineers to compare a building's energy performance against customized peer groups, identify energy performance opportunities, and set energy performance. It can also be used by energy performance program implementers to analyze energy performance features and trends in the building stock. The BPD compiles data from various data sources, converts it into a standard format, cleanses and quality checks the data, and provides users with access to the data in a way that maintains anonymity for data providers.

The BPD consists of the database itself, a graphical user interface allowing exploration of the data, and an application programming interface allowing the development of third-party applications using the data.

Performance comparison based on the 5-way alignment of five networks of size 1500, 2000, 2500, 3000 and 3000. The last two rows are obtained by considering only equivalence classes that contain at least one node from every species. The performance of each method is assessed using the following metrics: specificity(SP), number of correct nodes (CN), and mean normalized entropy (MNE). In each metrics, best performance is shown in bold.

The goal of this study was to adapt a recently proposed linear large-scale support vector machine to large-scale binary cheminformatics classification problems and to assess its performance on various benchmarks using virtual screening performance measures. We extended the large-scale linear support vector machine library LIBLINEAR with state-of-the-art virtual high-throughput screening metrics to train classifiers on whole large and unbalanced data sets. The formulation of this linear support machine has an excellent performance if applied to high-dimensional sparse feature vectors. An additional advantage is the average linear complexity in the number of non-zero features of a prediction. Nevertheless, the approach assumes that a problem is linearly separable. Therefore, we conducted an extensive benchmarking to evaluate the performance on large-scale problems up to a size of 175000 samples. To examine the virtual screening performance, we determined the chemotype clusters using Feature Trees and integrated this information to compute weighted AUC-based performance measures and a leave-cluster-out cross-validation. We also considered the BEDROC score, a metric that was suggested to tackle the early enrichment problem. The performance on each problem was evaluated by a nested cross-validation and a nested leave-cluster-out cross-validation. We compared LIBLINEAR against a Naïve Bayes classifier, a random decision forest classifier, and a maximum similarity ranking approach. These reference approaches were outperformed in a direct comparison by LIBLINEAR. A comparison to literature results showed that the LIBLINEAR performance is competitive but without achieving results as good as the top-ranked nonlinear machines on these benchmarks. However, considering the overall convincing performance and computation time of the large-scale support vector machine, the approach provides an excellent alternative to established large-scale classification approaches.

Performance comparison of various models.

The protein-protein interface comparison software PiMine was developed to provide fast comparisons against databases of known protein-protein complex structures. Its application domains range from the prediction of interfaces and potential interaction partners to the identification of potential small molecule modulators of protein-protein interactions.[1]

The protein-protein evaluation datasets are a collection of five datasets that were used for the parameter optimization (ParamOptSet), enrichment assessment (Dimer597 set, Keskin set, PiMineSet), and runtime analyses (RunTimeSet) of protein-protein interface comparison tools. The evaluation datasets contain pairs of interfaces of protein chains that either share sequential and structural similarities or are even sequentially and structurally unrelated. They enable comparative benchmark studies for tools designed to identify interface similarities.

Data Set description:

The ParamOptSet was designed based on a study on improving the benchmark datasets for the evaluation of protein-protein docking tools [2]. It was used to optimize and fine-tune the geometric search parameters of PiMine.

The Dimer597 [3] and Keskin [4] sets were developed earlier. We used them to evaluate PiMine’s performance in identifying structurally and sequentially related interface pairs as well as interface pairs with prominent similarity whose constituting chains are sequentially unrelated.

The PiMine set [1] was constructed to assess different quality criteria for reliable interface comparison. It consists of similar pairs of protein-protein complexes of which two chains are sequentially and structurally highly related while the other two chains are unrelated and show different folds. It enables the assessment of the performance when the interfaces of apparently unrelated chains are available only. Furthermore, we could obtain reliable interface-interface alignments based on the similar chains which can be used for alignment performance assessments.

Finally, the RunTimeSet [1] comprises protein-protein complexes from the PDB that were predicted to be biologically relevant. It enables the comparison of typical run times of comparison methods and represents also an interesting dataset to screen for interface similarities.

References:

[1] Graef, J.; Ehrt, C.; Reim, T.; Rarey, M. Database-driven identification of structurally similar protein-protein interfaces (submitted)
[2] Barradas-Bautista, D.; Almajed, A.; Oliva, R.; Kalnis, P.; Cavallo, L. Improving classification of correct and incorrect protein-protein docking models by augmenting the training set. Bioinform. Adv. 2023, 3, vbad012.
[3] Gao, M.; Skolnick, J. iAlign: a method for the structural comparison of protein–protein interfaces. Bioinformatics 2010, 26, 2259-2265.
[4] Keskin, O.; Tsai, C.-J.; Wolfson, H.; Nussinov, R. A new, structurally nonredundant, diverse data set of protein–protein interfaces and its implications. Protein Sci. 2004, 13, 1043-1055.

Performance comparison of machine learning models across accuracy, AUC, MCC, and F1 score on GMSC dataset.