User-centric Activity Recognition and Prediction Model using Machine Learning Algorithms
Open accessThe DOI matches its Crossref record; 30 of the 31 references it cites were found in a registry or an index; none of the 5 sources its claims cite is checked for support yet.
Abstract
Human Activity Recognition has been a dynamic research area in recent years. Various methods of collecting data and analyzing them to detect activity have been well investigated. Some machine learning algorithms have shown excellent performance in activity recognition, based on which many applications and systems are being developed. Unlike this, the prediction of the next activity is an emerging field of study. This work proposes a conceptual model that uses machine learning algorithms to detect activity from sensor data and predict the next activity from the previously seen activity sequence. We created our activity recognition dataset and used six machine learning algorithms to evaluate the recognition task. We have proposed a method for the next activity prediction from the sequence of activities by converting a sequence prediction problem into a supervised learning problem using the windowing technique. Three classification algorithms were used to evaluate the next activity prediction task. Gradient Boosting performs best for activity recognition, yielding 87.8% accuracy for the next activity prediction over a 16-day timeframe. We have also measured the performance of an LSTM sequence prediction model for predicting the next activity, where the optimum accuracy is 70.90%.
Overview
CoChat Summary
TL;DR
This paper proposes a conceptual user-centric model that uses machine learning to recognize activities from wrist-worn sensor data and predict the next activity from the recognized activity sequence. Gradient Boosting achieved 99.7% recognition accuracy, while next-activity prediction reached 87.8% for a 16-day supervised windowing model and 70.90% for an LSTM model.
Key findings
- For activity recognition, Gradient Boosting had the highest accuracy at 99.7%; KNN was 99.2%, Random Forest 99.3%, ANN 99.1%, SVM 98.7%, and Naïve Bayes 61.9%. Removing the three novel features (sound level, day-of-week, time-of-day) reduced all algorithms, with Gradient Boosting still best at 98.8% and Naïve Bayes lowest at 54.4%; the drop was around 2% for all algorithms except Naïve Bayes, which dropped almost 7%.
- For next-activity prediction using the windowing-based supervised approach, the 16-day timeframe outperformed the 8-day timeframe. On 16 days, Gradient Boosting reached 87.8% at window size 6, ANN 81.6% at window size 6, and KNN 82.7% at window size 6. On 8 days, Gradient Boosting reached 61.2% at window size 4, ANN 53.1% at window size 2, and KNN 46.9% at window sizes 2 and 3.
- The LSTM sequence prediction model achieved 68.75% accuracy for the 8-day timeframe with window size 6 and 70.90% for the 16-day timeframe with window size 8. For 8 days and window size 6, LSTM outperformed the best supervised model; for 16 days and window size 8, the supervised model performed better than LSTM (80.4% vs. 70.9%).
- The paper reports that both prediction methods perform better with a longer timeframe of activities.
Methods
- The recognition dataset contained 330,993 rows and ten columns, collected from three participants (two male, one female), each recording data for three consecutive days consisting of two weekdays and one weekend. Data were collected using the Android app AndroSensor on a smartphone tied to the wrist, recording accelerometer, gyroscope, sound level, and timestamps at 0.125 s per instance. Features included acceleration x/y/z, orientation x/y/z, sound level, day-of-week, and time-of-day; the last three were introduced as novel features.
- Activity recognition evaluated six algorithms: ANN, KNN, Naïve Bayes, Random Forest, SVM, and Gradient Boosting. Next-activity prediction used ANN, KNN, and Gradient Boosting, plus an LSTM sequence prediction model. The sequence prediction problem was converted into a supervised learning problem using a windowing technique; window sizes ranged from 2 to 6 for the 8-day prediction dataset and 2 to 8 for the 16-day dataset, while LSTM used fixed window sizes of 6 for 8 days and 8 for 16 days. The prediction dataset was collected from one participant over 8 and 16 consecutive days.
- Implementation used Python with scikit-learn, Keras and TensorFlow for LSTM, and Anaconda Jupyter Notebook (Version 3). Default hyperparameters were modified, and training/test accuracy plus a confusion matrix for the best recognition model were reported.
Limitations
- Stated: resource constraints prevented using a smartwatch or wearable device, so a smartphone tied to the wrist was used; data collection depended on only three participants; batch processing was used due to lack of high-end machines and components; accuracy was moderate considering the dataset and results were not always accurate; prediction accuracy depends mainly on timeframe, and it is difficult to say which model predicts next activity more accurately; future work plans more participants.
- Stated/evident: overfitting is indicated by training accuracy increasing with window size while test accuracy generally decreases, especially for smaller timeframe datasets.
- Evident: the prediction dataset came from only one participant, limiting user-centric generalizability, and the recognition dataset was small (three participants, three days each); the model was not deployed in real-time with streaming data.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 0 Not valid · 5 Not checked · 67 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We have noticed that both prediction models yield better accuracy with a larger window size when the time frame increases from 8 days to 16 days.”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Gradient Boosting performs best for activity recognition, yielding 87.8% accuracy for the next activity prediction over a 16-day timeframe.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We have also measured the performance of an LSTM sequence prediction model for predicting the next activity, where the optimum accuracy is 70.90%.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 2 assumptions not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Gradient Boosting performs best for activity recognition, yielding 87.8% accuracy for the next activity prediction over a 16-day timeframe.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Almost all the instances of the actual class have been predicted accurately by the model.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
42 more claims
Create a free account to see all 72 claims and every source verdict.
- Probably not a substantial claim. 6
- It may state two separate claims as one. 12
- The quote only partly states this claim. 7
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ III. ARCHITECTURE OF PROPOSED SOLUTION · p. 4
“time-of-the-day (Morning, Noon, Evening, Night).”
§ III. ARCHITECTURE OF PROPOSED SOLUTION · p. 4
“| Afternoon (17:00 PM- 19:59 PM) | 3 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ III. ARCHITECTURE OF PROPOSED SOLUTION · p. 5
“| 5 | 10 | 2 | 6 | 9 |”
§ III. ARCHITECTURE OF PROPOSED SOLUTION · p. 5
“| 1 | 6 | 4 | 5 10 | 2 | 6 | 3 1 | ….. |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ IV. RESULT AND DISCUSSION · p. 8
“the test accuracy started to decline and continued for all the algorithms.”
§ IV. RESULT AND DISCUSSION · p. 8
“| 5 | 98.2 | 37.5 | 99.1 | 39.6 | 95.5 | 58.3 | | 6 | 99.1 | 41.7 | 96.4 | 31.2 | 96.4 | 43.2 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ IV. RESULT AND DISCUSSION · p. 7
“With the increase of window size, training accuracy kept increasing.”
§ IV. RESULT AND DISCUSSION · p. 8
“| 7 | 93.8 | 71.1 | 96.9 | 74.2 | 96.9 | 81.4 | | 8 | 95.4 | 68.0 | 96.4 | 75.3 | 96.4 | 80.4 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ IV. RESULT AND DISCUSSION · p. 8
“But for GB, the accuracy increased up to window size 6, and after that, it started to decrease”
§ IV. RESULT AND DISCUSSION · p. 8
“| 4 | 86.4 | 64.3 | 87.2 | 71.4 | 89.0 | 81.6 | | 5 | 92.9 | 70.4 | 95.6 | 76.5 | 95.6 | 78.6 |”
Citations
Cites
Scikit-learn: Machine Learning in Python
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Müller, A., Nothman, J., Louppe, G., et al. (2012) · arXiv · doi:10.48550/arxiv.1201.0490 · Cited by 71,016
Scikit-learn: Machine Learning in Python
(2011) · Journal of Machine Learning Research · doi:10.5555/1953048.2078195 · Cited by 8,539
Activity Recognition from User-Annotated Acceleration Data
Bao, L., & Intille, S. S. (2004) · Lecture Notes in Computer Science · doi:10.1007/978-3-540-24646-6_1 · Cited by 3,147
Swan, M. (2012) · Journal of Sensor and Actuator Networks · doi:10.3390/jsan1030217 · Cited by 934
Machine Learning Methods for Classifying Human Physical Activity from On-Body Accelerometers
Mannini, A., & Sabatini, A. M. (2010) · Sensors · doi:10.3390/s100201154 · Cited by 781
How smart are our environments? An updated look at the state of the art
Cook, D. J., & Das, S. K. (2007) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2006.12.001 · Cited by 589
Human activity prediction: Early recognition of ongoing activities from streaming videos
Ryoo, M. S. (2011) · 2011 International Conference on Computer Vision · doi:10.1109/iccv.2011.6126349 · Cited by 556
Vision-based human activity recognition: a survey
Beddiar, D. R., Nini, B., Sabokrou, M., & Hadid, A. (2020) · Multimedia Tools and Applications · doi:10.1007/s11042-020-09004-3 · Cited by 505
Complex Human Activity Recognition Using Smartphone and Wrist-Worn Motion Sensors
Shoaib, M., Bosch, S., Incel, O., Scholten, H., & Havinga, P. (2016) · Sensors · doi:10.3390/s16040426 · Cited by 423
Human Activity Recognition from Accelerometer Data Using a Wearable Device
Casale, P., Pujol, O., & Radeva, P. (2011) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-21257-4_36 · Cited by 335
Research works printed but not linked: 3
- [3][3] R. Rawassizadeh, B. A. Price, and M. Petre, "Wearable devices: Has the Age of Smart watches Finally Arrived?," Communications of the ACM, 2015, pp. 45-47. -
- [16][16] N. C. Krishnan, D. Colbry, C. Juillard and S. Panchanathan, "Real time human activity recognition using tri-axial accelerometers, Sensors, signals and information processing workshop," Sensors Signals and Information Processing Workshop, Sedona, AZ, 28,2018, pp.915-922. -
- [19][19] H. Wang and J. Wu, "Classification of Human Posture and Movement Using Accelerometer Data," in International Conference on Innovations in Computing & Networking (ICICN16), CSE, RRCE, ISSN: 09750282. -
Cited by
Predicting reaction based on customer's transaction using machine learning approaches
Hayder, I. M., Al-Ali, G. A., & Younis, H. A. (2022) · International Journal of Power Electronics and Drive Systems/International Journal of Electrical and Computer Engineering · doi:10.11591/ijece.v13i1.pp1086-1096 · Cited by 16
Transformer Based Ensemble Framework For Sequential User Behavior Prediction
Boyapati, A., Sinha, P., Rayavarapu, A., & Sreepada, R. (2024) · Conference on AI, Science, Engineering, and Technology (AIxSET) · doi:10.1109/aixset62544.2024.00045 · Cited by 1
Metadata
- Authors
- Roy, N., Ahmed, R., Huq, M. R., & Shahriar, M. M.
- Affiliations
- East West University (BD)
- Venue
- International Journal of Advanced Computer Science and Applications
- Published
- 2021
- Volume
- 12
- Issue
- 12
- ISSN
- 2156-5570 (ISSN-L)
- 2158-107X
- OpenAlex
- W4206749371
- Cited by
- 2
- References
- 31
- Citations by year
- 2024: 1 · 2022: 1
- FWCI
- 0.18
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- Locations
- International Journal of Advanced Computer Science and Applications (publishedVersion, cc-by)
- International Journal of Advanced Computer Science and Applications (submittedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Social Sciences › Transportation › Human Mobility and Location-Based Analysis
- Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
- Keywords
- human activity recognition, next activity prediction, sensor data, window method, gradient boosting, LSTM, user-centric systems, activity recognition dataset
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: User-centric Activity Recognition and Prediction Model using Machine Learning Algorithms. https://app.cochat.ai/papers/10.14569/ijacsa.2021.0121265
- The paper
- Roy, N., Ahmed, R., Huq, M. R., & Shahriar, M. M. (2021). User-centric Activity Recognition and Prediction Model using Machine Learning Algorithms. International Journal of Advanced Computer Science and Applications. https://doi.org/10.14569/ijacsa.2021.0121265
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.